anderson localization in two-channel wires with · bio-molecule called the dna (deoxyribo-nucleic...
TRANSCRIPT
APPROVED: Arkadii Krokhin, Major Professor Paolo Grigolini, Committee Member Arup Neogi, Committee Member William D. Deering, Committee Member Christopher Littler, Interim Chair of the
Department of Physics Sandra L. Terrell, Dean of the Robert B.
Toulouse School of Graduate
ANDERSON LOCALIZATION IN TWO-CHANNEL WIRES WITH
CORRELATED DISORDER: DNA AS AN APPLICATION
V. M. Kemal Bagci
Dissertation Prepared for the Degree of
DOCTOR OF PHILOSOPHY
UNIVERSITY OF NORTH TEXAS
December 2007
Bagci, V. M. Kemal. Anderson Localization in Two-Channel Wires with
Correlated Disorder: DNA as an Application. Doctor of Philosophy (Physics),
December 2007, 66 pages, 27 illustrations, references, 104 titles.
This research studied the Anderson localization of electrons in two-channel
wires with correlated disorder and in DNA molecules. It involved an analytical
calculation part where the formula for the inverse localization length for electron
states in a two-channel wire is derived. It also involved a computational part where
the localization length is calculated for some DNA molecules.
Electron localization in two-channel wires with correlated disorder was studied
using a single-electron tight-binding model. Calculations were within second-order
Born-approximation to second-order in disorder parameters. An analytical
expression for localization length as a functional of correlations in potentials was
found.
Anderson localization in DNA molecules were studied in single-channel wire
and two-channel models for electron transport in DNA. In both of the models, some
DNA sequences exhibited delocalized electron states in their energy spectrum.
Studies with two-channel wire model for DNA yielded important link between electron
localization properties and genetic information.
ii
Copyright 2007
by
V. M. Kemal Bagci
iii
ACKNOWLEDGEMENTS
I foremost thank my supervisor Prof. Arkadii Krokhin, for his support and
valuable discussions. He introduced me to the topic of disordered systems and has
enlarged my understanding of physics. I thank my wife Eylem for her support and
invaluable presence during my studies. I would also like to thank my family for their
encouragement and support.
I would like to thank Prof. Paolo Grigolini for discussions on the topic. I would
also like to extend my thanks to Prof. Arup Neogi and Prof. Bill Deering of our
department for their interest in my studies.
I would like to thank the staff of our physics department, especially Kari for
being such a big help.
I will also always remember the cheerful comments of Prof. Samuel Matteson.
I thank my friend Krista for her help and support.
This work was partially supported by DOE.
iv
TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS ............................................................................................... iii LIST OF FIGURES.......................................................................................................... v Chapters 1. INTRODUCTION.................................................................................................. 1
1.1 Structure of DNA........................................................................................ 4 2. THE CONCEPT OF LOCALIZATION ................................................................... 8
2.1 Development of Localization Theory, Metal-Insulator Transition and Scaling Theory of Localization ................................................................. 12
2.2 Beyond Scaling Theory and Localization of 1-D Wires ............................ 17 2.3 Analytical Results for Localization Length................................................ 19
3. CALCULATION OF ELECTRON LOCALIZATION LENGTH FOR TWO-
CHANNEL WIRES.............................................................................................. 25 3.1 Transfer-Matrix Formalism....................................................................... 27 3.2 Lyapunov Exponent ................................................................................. 31 3.3 Scattering Matrix and Conductance......................................................... 32 3.4 Calculation of Transfer Matrix in Second-Order Born Approximation ...... 34 3.5 One Evanescent and One Propagating Mode ......................................... 38 3.6 Discussions.............................................................................................. 41
4. ELECTRON LOCALIZATION IN DNA ................................................................ 45
4.1 Fishbone Model of DNA........................................................................... 46 4.2 Numerical Calculations ............................................................................ 49 4.3 Two-Chain Model of DNA ........................................................................ 51 4.4 Gene Expression and Quantum Electron Transport ................................ 53
5. CONCLUSIONS ................................................................................................. 60 REFERENCES.............................................................................................................. 63
v
LIST OF FIGURES
Page
1.1 Chemical Structure of DNA .............................................................................5
1.2 Graphical Depiction of the Structure of DNA ...................................................6
2.1 Localization of Water Waves .........................................................................12
2.2 Localized and Extended States in a Disordered Medium ..............................13
2.3 Scaling Parameter vs Natural Logarithm .......................................................16
2.4 Anderson Tight-Binding Model of a 1-D Chain ..............................................19
2.5 Sharp Mobility Edge in Energy Spectrum ......................................................21
2.6 Microwave Transmission Experiment Setup..................................................22
2.7 Microwave Transmission Experiment Results ...............................................23
2.8 Anderson Tight-Binding Model for a 2-D Chain .............................................23
3.1 Slice of a Two-Chain Wire .............................................................................30
3.2 Comparison of Results for Localization Length in Two-Channel Wires .........42
4.1 Fishbone Model for DNA ...............................................................................47
4.2 Fishbone Model Dispersion Curves...............................................................48
4.3 Binary Correlation Function of BRCA Sequence ...........................................49
4.4 Binary Correlation Function of Chromosome 22............................................50
4.5 LL for BRCA Gene (Fishbone Model)............................................................51
4.6 LL for Chromosome 22 (Fishbone Model) .....................................................52
4.7 LL for BRCA (Two-Channel Model) ...............................................................54
4.8 LL for Chromosome 22 (Two-Channel Model)...............................................54
4.9 LL for SNAP29 Exon 5 ..................................................................................55
4.10 LL for SNAP29 Intron 2 .................................................................................56
vi
4.11 LL for the Whole SNAP29 Gene....................................................................56
4.12 LL for SYLB1 Exon 7 .....................................................................................57
4.13 LL for SYLB1 Intron 2 ....................................................................................58
4.14 LL for the Whole SYLB1 Gene ......................................................................58
4.15 Correlation Function for SNAP29 Exon 5 ......................................................59
4.16 Correlation Function for SNAP29 Intron 2 .....................................................59
CHAPTER 1
INTRODUCTION
The search for the secret of life is perhaps the most intriguing and rewarding intellectual
adventure of all human endeavors. Countless ideas and ideologies have fought over the
ultimate answer throughout history. The secret is well hidden, not only because of the
in�nite delicacy and intrigues of what we may call life, but also the fog of time that divides
the questionable origin and its mechanism. At this point in science, we have not yet fully
put forward a scienti�c theory for the origin of life on Earth, but we have accumulated a
signi�cant amount of information on how it carries on.
So far all life forms we have found originate from another living organism or organisms,
hence they borrow a number of characteristic traits that we name as genetic information from
their ancestors. The genetic information organisms pass to their o�spring is such a valuable
asset that its continuity is a main driving force behind many animal and human behaviors. It is
only recent that we have established an understanding on how they pass the information and
we are still trying to understand the extent of the e�ect of that information on an organism's
life span. We have also found out that the genetic information passed on is not always an
exact replica or a mix of both ancestors, but that it may show slight variations from its source.
Furthermore, we found that such variations may occur in the body of an organism during its
lifetime, sometimes leading to fatal diseases.
The genetic information is stored in all living cells of an organism, in form of a very long
bio-molecule called the DNA (deoxyribo-nucleic acid). Nucleic acids were �rst isolated in
1869 by Friedrich Miescher, but it took almost 75 years to demonstrate that DNA was the
carrier of genetic information. The structure of DNA was discovered by James Watson and
Francis Crick in 1953 [1]. It is a polymer of nucleotides, which consist of phosphorylated
sugars and bases. The bases are adenine, thymine, cytosine and guanine. The nucleotides
1
are arranged in a two-stranded one-dimensional chain. The structure of DNA will be discussed
furthermore in the text.
The genetic information within DNA is stored as a sequence of nucleotides. This is a very
long sequence and it varies between di�erent species and within the members of the same
species. The sequence of nucleotides typically look like ...ATCGATTGCCACTTAGCAGG...
(for a single strand). The relationship between the nucleotide sequences of genes and the
amino-acid sequences of proteins is determined by the rules of translation, known collectively
as the genetic code. The genetic code consists of three-letter 'words' called codons formed
from a sequence of three nucleotides (e.g. ACT, CAG, TTT). These codons are signals for
the ribosome (protein factory of a cell) and various enzymes to bind a speci�c amino-acid
(building blocks of proteins) with other amino-acids, in relation to the following or preceding
codon in the sequence. While there are 64(43) possible codons, there are only 24 di�erent
amino-acids in nature, so some codons point at the same amino-acid.
DNA has regions of di�erent functions. It contains large segments that do not play any
role in the transcription (protein coding). These regions are called introns. The regions that
are expressed during transcription (sequences that lead to protein coding) are called exons.
Along the DNA sequence, there are regions that signal or mark the beginnings of transcrip-
tion regions. There are sequences that will stop transcription, sequences that enhance the
transcription, and sequences that control or regulate the gene expression. There are volumes
of studies in almost every property of these various regions of DNA.
The sequence of nucleotides is a random sequence, in the sense that which nucleotide
is the next in chain can not be known from previous order. However, the sequences do not
appear to be totaly uncorrelated. In fact, the DNA sequences display long-range correlations
if we focus on parts of DNA. Whether there is a di�erence between the statistics of intron
and exon regions of DNA has also been investigated thoroughly [2, 3, 4, 5, 6]. There is still
a wide open and rich �eld of study for physics to �nd out the relationships between physical
properties and statistical properties of DNA.
2
The electronic structure of DNA poses a very interesting problem by itself. Surely the mo-
bility of electrons along the sequence e�ects DNA's functional behavior (signalling, enzyme-
DNA interactions etc.) and its structural-sequential integrity, we just don't know how. The
charge migration is believed to be important for the radiation damage repair [7]. One re-
cent study has shown that single-base mutations on DNA molecules may lead to signi�cant
changes in conductance [8]. DNA double helices are also expected to be particularly use-
ful for molecular electronic devices [9, 10, 11]. Numerous studies on conductance of DNA
yielded results from insulator, to semi-conductor or metallic [9, 12, 13]. Although it is well-
understood now that the current in DNA is due to light particles, i.e. electrons or holes
[14], the physical mechanism that provides a wide variety of conducting (or non-conducting)
properties for di�erent DNA is still unclear. The electrical conductivity of the DNA molecules
exhibits strong environmental dependence that includes variation with temperature, chemical
composition of the solution, humidity, quality of the metallic contacts, etc [15]. It is not clear
whether these properties play a role in living nature or whether they are just artifacts of the
experiments.
Within the statistical nature of DNA, there lies an important clue on the electronic behav-
ior along the molecule, namely quantum localization of electrons (or holes). The quantum
mechanical charge carrier has a wavefunction that extends to a certain spatial region. The
extent of the wavefunction depends on the type of motion the electron executes, which in
return depends on its total energy and the potential function it moves under the in uence
of. A free electron, for example, extends to in�nity. A Gaussian wave packet will spread
in�nitely in free space. An electron in an ordered system, where potential function exhibits
a periodic nature, will extend throughout the whole system (delocalized quantum state) if
the electrons energy is within the allowed zone. It is the disordered, random systems that
exhibit the crucial property of electron localization, which is the case when the extend of the
electrons' wavefunction, the localization length, is �nite and the electron is exponentially lo-
calized within a �nite region. The localization length determines the conductance properties
and electron-ion (molecules or enzymes) interactions for DNA. One dimensional disordered
3
systems like DNA were considered to have only localized states [16, 17]. However it has been
shown that for one-dimensional single-channel wires, a group of systems which DNA is often
modeled within, delocalized states may exist under certain conditions [18, 19]. It seemed
also probable that delocalized states may exist for one-dimensional two-channel wires, which
is a more accurate approximation for DNA. In this thesis, the electron localization lengths
for DNA samples in two di�erent tight-binding models (the single channel Fishbone model
and the two-channel wire) were calculated. In order to apply the two-channel model, it was
necessary to �nd an analytical expression for the localization length for those systems, taking
into account all correlations. That work is also presented in the thesis.
Our knowledge of how DNA exactly functions inside a living organism is limited; indirect
observations, snapshots and in vitro observations are the main source of information about
the mechanism of the expression of information carried in DNA [20]. Since it is a low-
dimensional system with small extend in constrained dimensions, the problem of electronic
behavior in DNA is a quantum mechanical one in its nature. Ab-initio studies for single
nucleotides or short DNA molecules, or for molecules with periodic repetitions (arti�cal DNA
molecules) can be carried out [21, 22, 23], however it is not possible within today's technology
to study DNA in an exact manner. Instead, we rely on statistical and approximate methods.
In this thesis, the DNA molecules are studied using one of such approximate models, the
so-called tight-binding model. A detailed discussion on the model follows in the text.
1.1. Structure of DNA
Deoxyribo-nucleic acid, DNA, is a very long chain of nucleotides, phosphorylated sugars
attached to bases adenine, thymine, cytosine and guanine. The bases are organic compounds
(Fig.1.1).
One nitrogen from each base bonds with a 2'-deoxyribose molecule and they form the
nucleotide. Then the 3' hydroxyl group of one nucleotide sugar makes a phosphodiester bond
with the 5' phosphate group of the consecutive one as shown in Fig.1.1. The next base
in the sequence is then linked to this phosphorylated sugar [24]. The polynucleotide chain
has a sense of direction with one end being the 5' phosphate group and the other the 3'
4
Figure 1.1. Chemical structure of DNA bases and the sugar-phosphate group
(yellow and brown) is shown above.
hydroxyl group. For each chain like this, there is a complementary one with the opposite
sense of direction. The two chains are linked only through the matching bases; adenine
always pairs with thymine (two hydrogen bonds) and guanine with cytosine (three hydrogen
bonds) (Fig.1.2). Contrary to the hydrogen bond formation in base pairs of A-T and C-G,
all other pairings, i.e. A-G or C-T, lead to a signi�cant repulsive electrostatic potential. The
phosphorylated sugars form the backbone of DNA. The electrostatic interactions between
the negatively charged backbone and the positively charged bases, and the hydrogen bonds
between the hydrophobic bases hold the DNA together.
The length of a DNA molecule can change anywhere between 106 to 108 base pairs (A-T
or C-G). The total length of human DNA molecules (for a single cell) is around 3 billion base
pairs. These molecules exist in a tightly wrapped structure inside a living cell for most of the
5
Figure 1.2. Structure of DNA is shown. a) The double-helix structure, the
four bases A, T, C and G shown in their complimentary form. b) Graphical
depiction of the chemical structure.
time, until they are to be used for replication or protein coding, at which point they unfold.
The extended DNA resembles a helical ladder, in which the phosphorylated sugars form the
side rails. When they are idle, they are found in forms of chromosomes, long chains of DNA
molecules wrapped around histone proteins.
The diameter of the helix is 20 �A and the vertical distance between consecutive bases 3:4
�A. One turn of the helix consists of 10 bases and the twist angle is normally 36o . The values
for distances are taken to be the same when the helical structure is unfolded into a ladder
structure.
The �-orbitals of adjacent bases, on the same chain or across, are all perpendicular to
the length of the DNA strands. These orbitals overlap signi�cantly, creating what is called �-
stacking, lowering the total energy of the system and further stabilizing the DNA molecule.
The overlap interaction between bases forming base pairs are stronger in comparison to
6
the overlap interaction between adjacent (consecutive bases along a strand) bases, for two
reasons. Firstly, the distance between bases on the same strand is larger then the distance
between bases on complimentary strands. Secondly, the �-orbitals of bases on complimentary
strand have the same direction, while there is a tilting of orbital directions (a 36o turn at
each consecutive base pair) for bases along the same strand in the helical structure [99].
7
CHAPTER 2
THE CONCEPT OF LOCALIZATION
A crystal is a perfect periodic structure, where every physical property along the directions
of translational symmetry is identical. For example, the possibility of �nding an electron is
the same for each unit cell. A crystal structure is not necessarily three-dimensional. A chain
of identical atoms with equal spacings between them is an example of a one-dimensional
crystal. The electrons inside the crystal that are not bound to the ions or molecules are called
crystal electrons and they occupy quantum states that satisfy the periodicity conditions of
the structure. The eigenfunctions of periodic systems are called Bloch wavefunctions. If the
potential function of the system has a periodicity
(2.1) V (r) = V (r + T);
the Bloch wavefunction has the form
(2.2) k(r) = uk(r) exp({k � r)
with uk(r) = uk(r + T). The wavefunction then satis�es
(2.3) k(r + T) = k(r) exp({k � T); j k(r + T)j2 = j k(r)j2
and each k spreads throughout the whole system. The energies of the wavefunctions
are E(k), smooth functions of the wavevector k. At the absolute temperature T = 0 K, the
crystal electrons occupy the eigenstates beginning from the one with the lowest energy until
the electrons are depleted. There are no excitations at that temperature, thus eigenstates
solely dictate the electronic transport in the crystal. While all crystal wavefunctions are
8
delocalized, that is they are spread out to the whole crystal, this does not mean that they
contribute to conductance. The conductivity of a material is much more complex a property
that depends on the whole eigenstate structure. The presence of delocalized states is a
necessary but not su�cient condition for conductivity.
The physics of periodic systems is well formulated. However, it is extremely unlikely to
�nd perfect periodicity even for arti�cial materials. There exist deviations from the perfect
periodic structure. These deviations are called defects or impurities. A simple defect may be
an atom B at a single site instead of having atom type A at every lattice point (the collection
of all points that form the crystal). Similarly, one of the atoms A may shift slightly from its
position, or may not be there at all.
� ...A...A...A...A...A...A...A...A... Perfect crystal
� ...A...A...A...B...A...A...A...A... Atom A replaced by atom B
� ...A...A...A... ...A...A...A...A... Missing atom (vacancy)
� ...A...A..A....A...A...A...A...A... Shift of one of the atoms, topological defect
� ...A...A..A........A...B...A...B... Multiple defects
� ...A...B...C...A...C...B...B...A... Topological periodicity, quasi-periodic structure
� ...A...B..E...B.C.....E...A..A..C.. No periodicity or order, complete disorder
Putting aside atomic or molecular scale objects, many material mediums in nature do not
have a periodic structure at all. However for most physically, technologically or biologically
important materials, it is possible to treat the material medium as quasi-periodic or crystal-
like and consider randomness or disorder in the medium as defects. Any system without a
perfect periodicity is considered a disordered system. The nature of individual defects and
the distribution of these defects determine the strength of disorder. The disorder strength
can vary between very weak disorder in crystal-like structures with infrequent defects (e.g.
silicon wafers) to complete disorder (e.g. amorphous silicon).
When the disorder is not signi�cant, the physical characteristics of the disordered sys-
tem can be calculated by perturbation theory taking the perfectly periodic structure as the
9
zeroth-order basis. Transport properties and conductivity are among the successfully cal-
culable characteristics. The Bloch-Boltzmann quasiclassical theory of electronic transport in
lightly disordered conductors has been quite useful in describing the impurity and temperature
dependencies of conductance in ordinary relatively pure conductors [25]. Further transport
properties such as magnetoresistivity, Hall e�ect, and thermal conductivity can also be han-
dled with success. However the perturbation theory fails at a point with increasing strength
of disorder. The weak-scattering theory can no longer explain phenomena like the increase in
resistance with decreasing temperature in disordered dirty systems [26, 27, 28]. The concept
of localization comes in to help in understanding strong disorder.
Localization is a wave property. Electromagnetic waves, water waves or particle waves
all can display localization. The phenomena of localization exists when waves interact with
disordered mediums. Due to disorder, there are many scattering centers inside the medium.
Waves undergo a number of multiple scattering processes when they propagate through the
material. Let's consider as an example an electron undergoing a multiple scattering process
from an initial state jkii with a wavevector ki into the �nal state jkfi with the wavevector kf .
The scattering process PA � (ki ! kA;1 ! kA;2 ! � � � ! kA;n�2 ! kA;n�1 ! kf) represents
one path involving n scattering events and has a probability amplitude TA. A similar process
PB � (ki ! kB;1 ! kB;2 ! � � � ! kB;n�2 ! kB;n�1 ! kf) might exist with probability
amplitude TB. In general, the alternative paths accumulate di�erent phases, �A and �B. The
probability of transition jkii ! jkfi is then given by
(2.4) jTA + TBj2 = jTAj2 + jTBj2 + 2jTAjjTBj cos(�A � �B):
Since in a disordered system, the phases �A and �B are random, the ensemble average
hcos(�A � �B)i = 0. So the transmission probability is jTAj2 + jTBj2, addition of probabilities
of individual processes. This simple addition of probabilities is valid for classical particles, i.e.
interference does not contribute. Let's now consider when jkfi = j � kii, which is the case
of backscattering. For a system with time-reversal invariance, the probability amplitude TA
10
of process PA and probability amplitude TB of process PB which correspond to time-reversed
process PA are equal, that is TA = TB. The phases �A and �B for the time-reversed trajecto-
ries are also equal. Then the transition probability for jkii ! j � kii is jTA + TBj2 = 4jTAj2[29]. Thus, we have an enhanced probability for backscattering (scattering in the opposite
direction). This phenomena is called coherent-backscattering. The wave interference be-
tween paths A and time-reversed A is always constructive independent of disorder, and can
completely halt the waves.
This absence of di�usion of waves in random medium is called Anderson-localization. The
localization of an electron implies that its wavefunction vanishes exponentially away from the
center of localization. The localization length is the measure of the spatial extension of the
localized state. Qualitatively,
(2.5) (x) � exp(� jx jl(E)
)
where l(E) is the localization length. The localization length, being a characteristic of a
given electronic eigenstate, depends on eigenenergy E.
If the localization length considerably exceeds all other relevant lengths in the system (size
of the system, mean-free path etc.), then the electron can extend to in�nity and contribute to
electron transport. The semi-classical Drude conductivity is slightly modi�ed due to coherent-
backscattering in this case and the correction to the conductivity is called weak-localization
correction. If l(E) is smaller in comparison to the system length, the particles occupying
those states do not interact with the boundaries, thus they do not contribute to electron
transport.
If all electrons occupy localized states at absolute zero temperature T = 0 K the dc-
conductivity vanishes and the system is insulating at that temperature. If there are electrons
in delocalized states, a non-zero dc-conductance may exist and the system may be conducting
at absolute zero temperature. On the other hand, at temperatures above absolute zero some
11
Figure 2.1. Localization is a wave phenomena. Here localization of water waves
in a bath is displayed. a) Crystal like structure. b) The disordered system with
localization of waves.
localized electrons may get excited to delocalized states and the system might conduct again.
For the rest of this thesis, all calculations and considerations are carried out at T = 0 K.
2.1. Development of Localization Theory, Metal-Insulator Transition and Scaling Theory of
Localization
2.1.1. Pre-Scaling Era
Anderson (1958) studied the electronic di�usion in a three-dimensional tight-binding
model in which the energies of the di�erent sites (lattice points) are randomly assigned from
a given interval according to a constant probability distribution with width W . He considered
a constant hopping parameter V between sites. He showed that above a critical value of
disorder, which he quanti�es as W=V , all states are spatially localized [16]. The relevance of
localization with regard to electronic transport properties of amorphous semiconductors was
discussed by Mott [30].
12
Figure 2.2. Electron states are localized or extended according to their location
with respect to the Fermi energy in the energy scale. If the Fermi energy lies in
the vicinity of the critical energy, the MIT may occur. The asymptotic behavior
of the system in that region forms the basis of one-parameter scaling theory.
The localized states and extended states cannot coexist at the same energy. Quantum
mechanical tunneling would admix them and destabilize the localized states [29]. Thus,
the localized states must be separated energy wise from the continuum of extended states,
without implying any gap in the density of states. The set of localized and delocalized states
are separated by a sharply de�ned energy Ec , which is called the critical energy or the mobility-
edge. The realization of the random disorder in a medium determines the position of mobility
edge. As the Fermi energy is tuned across the critical energy when an external parameter is
changed, the system goes through a transition between metallic and insulating phases. This
transition is called metal-insulator transition (MIT) or Anderson transition. A quantitative
approach to the MIT and calculations for mobility edge may be found in Thouless's early
work [31].
The presence of a mobility edge in disordered systems explains anomalous increase in
resistivity for dirty conductors. when T ! 0. At high temperatures, motion of electrons
in delocalized states are hindered by scattering from other crystal degrees of freedom, like
phonons. As the temperature drops, the number of phonons or other scatterers get smaller
so the mobility of the electrons increase, which in turn causes decrease in resistance. However
as temperature gets closer and closer to absolute zero, the electrons excited from localized
states to delocalized states start falling back to localized states. This reduces the number of
13
free carriers and increases resistance. This increase in resistance is called weak-localization
peak.
2.1.2. Single-Parameter Scaling Theory
Scaling theory of localization is a statistical approach to the problem of electron local-
ization in the vicinity of metal-insulator transition region. The so called critical region hosts
the states that contribute to the transport mechanisms of the disordered system. The study
of localization-delocalization transition as a function of energy across the mobility edge for
individual quantum states is too �ne-grained. We may however look at a physical quantity
that will give information about states at or around mobility edge. That quantity is electrical
conductance, which is the inverse of electrical resistance.
Let's at this point risk a deviation from the line of thought and focus on the conceptual
framework that allows us to investigate the conductance as a statistical variable. Conven-
tional statistical physics associates experimentally meaningful quantities with averages over
a statistical ensemble of macroscopically di�erent systems. For example, for a given macro-
scopic concentration of impurities there is a continuum of possible arrangements of their
positions in the host crystal. Although the results of measurements of a physical quantity,
when performed on speci�c members of the statistical ensemble, will be di�erent and de-
pendent on the speci�c con�guration of the impurities, or, more generally, on the speci�c
realization of the disorder, the statistical uctuations of the results will become vanishingly
small when compared with, say, the ensemble average, provided the system is su�ciently
large, i.e. macroscopic.
Physical quantities are usually assumed to ful�l this criterion, namely that they e�ectively
do not uctuate within the statistical ensemble of macroscopically equivalent systems in the
thermodynamic limit and such quantities are called self-averaging. Formally, the ensemble
average of a self-averaging quantity and the most probable value within the ensemble prac-
tically coincide when the size of the system is assumed to be in�nite. Self-averaging implies
that, in practice, measurements done on speci�c samples-speci�c realizations of the disorder
can be described in terms of ensemble averages.
14
Let's take a set aL of dimensionless physical parameters that vary with the system size
Ld , where L is to be considered as e�ective system size such as the inelastic scattering length
Li . We de�ne a parameter �,
(2.6)daLd lnL
= �(aL)
The important property of equations like Eq.2.6 is that for increasing L, the set aL goes
towards a simple subspace, often a line or even a point. Such points a? are known as
attractors. These points are de�ned by the condition �(a?) = 0. The set of parameters aL
can be reduced to a single parameter, named as the scaling parameter.
Now we shall consider once more the behavior of conductance of disordered systems with
its Fermi energy lying in the vicinity of the critical energy. If a scaling parameter related to the
quantum transport properties of disordered system is chosen, a thorough investigation of that
parameter will shed light on related properties like electron localization. Abrahams, Anderson,
Licciardello and Ramakrishnan [40] have come up with ln g as that scaling parameter, where
g is the dimensionless conductance. Their reasoning is simple : at the asymptotic limits of
g >> 1 (metallic regime) and g << 1 (insulator regime), the functional dependence of ln g
to system size L is independent of the size of the system. If we approach the mobility edge
either from the metallic regime or from the insulator regime, ln g can be approximated with
the corresponding asymptotic expression. They match each other at the point of transition.
The asymptotic expressions for the scaling parameter � are
(2.7) � � d ln gd lnL
= f d � 2;
const + ln g
g >> 1
g << 1:
The upper equality (2.7) derives from the Drude conductance formula for weakly disorder
metals. The lower equality follows from
(2.8) g � exp (�L=l(E)) ; L � �l(E) ln g
15
Figure 2.3. Scaling parameter � is a functional of the natural logarithm of the
dimensionless conductance. For � > 0 the behavior is metallic, whereas for
� < 0 it is insulating.
Finally, placing g from Eq.(2.8) into Eq.(2.7), the self averaging quantity-localization
length l(E) can be found from
(2.9)1l(E)
= � limL!1
12Lhln gi
where hln gi denotes the ensemble averaging, which the scaling theory predicts to reach
a limiting value for L!1.
A very famous result of the scaling theory of localization is the absence of extended states
in 1-D and 2-D systems. Since the upper limit for �(ln g) is given by d � 2 in the metallic
regime, it is clear that for d � 2, the value of � < 0 for all disorder realizations. This
means there is no attractor point and the system is always in localized regime. For 3-D
systems however, there exists a condition for delocalization, which depends on the strength
of disorder or the values of perturbation parameters with which the behavior of dimensionless
conductance is approximated around MIT point. Good and detailed explanation of these
16
points can be found in the review paper by Kramer and McKinnon [38] or in the excellent
books of Imry [25] and of Mello and Kumar [29].
The experimental work on the localization of electrons or holes in disordered systems have
been a vibrant �eld of study, and has also seen a revitalization along with the theoretical work
recently. The previously mentioned works of Mott [39], Wiesmann [26] , Mooij [27] and
Gorkov [28] were the pioneering studies into this �eld.
Application of a magnetic �eld leads to ux-dependent Aharonov-Bohm phase of a wave-
function, thus producing pronounced changes in the interference of the quantum mechanical
transport paths. This was most strikingly demonstrated a decade ago when the experimental
discovery of the Aharonov-Bohm-like oscillations of the magnetoresistance of thin normally
metallic cylinders provided direct evidence for the existence of quantum interference in the
presence of disorder [44].
If quantum interference is the dominant mechanism for the localization of states in a
random medium, localization e�ects should be of importance in other wave phenomena, too
[45]. That this is indeed the case has been demonstrated experimentally in recent years by
light scattering experiments [43, 46] and microwave localization [87].
2.2. Beyond Scaling Theory and Localization of 1-D Wires
In 1961 Mott and Twose [47] showed for a particular 1-D model, consisting of square
barriers of the same height and uctuation distance between the barriers, that all electronic
states are exponentially localized. In 1963, Borland reached the same conclusion for a 1-
D chain composed of equal potential barriers of �nite width and arbitrary shape separated
by zones of zero potential assigned randomly according to a continuous distribution [48].
However Tong later speci�ed that for Borland model extended states could exist for any
structural disorder if the individual potential unit exhibits resonances of the transmission [49].
More examples of one-dimensional random potentials is given in the review paper by Ishii [50].
The predictions of scaling theory and naive extension of the properties of the 1-D Anderson
model to other potentials have led to the belief that all the states in a random one-dimensional
system are exponentially localized. However, further research has shown that delocalized
17
states may exist. Deterministic quasi-periodic potentials were considered that can generate
localized or extended states depending on their parameters [51, 52].
In the models so far considered, the potentials were uncorrelated, i.e. white-noise like.
Dunlap [53] and Flores [54] initiated the consideration of short-range correlations in random
potentials. The short-range correlations may lead to isolated extended states in the spectrum
of the system. The random-dimer model (RDM) of Dunlap and his co-workers was used to
explain the high conductivity anomalies observed in certain organic polymers that should be-
have as insulators [55]. The RDM model has been extensively treated both from a theoretical
viewpoint [56, 57, 58, 59, 60] and it was experimentally realized in random semiconductor
superlattices [61, 62]. Other models like diluted Anderson-model [63, 64, 65], symmetrical
impurities in a pure chain [66], short-range correlations with classical analogues [67, 68] and
o�-diagonal disorder [69, 70] have also been studied.
The role of long-range correlations has also been analyzed during the last decade. Moura
and Lyra considered a 1-D tight-binding disordered model with 1=f -noise type long-range cor-
relations in the sequence of site-energies [80, 81]. They observed numerically the emergence
of a continuum of extended states and mobility edges for the carriers, marking the transition
between phases of localized states and extended states. Later on, Izrailev and Krokhin [18]
established an analytical relation between localization length and potential pair correlator and
showed how long-range correlations lead to the appearance of mobility edges. Their result is
valid in the second-order approximation over disorder. Though Tessieri [82] later showed that
to fourth-order approximation, the extended states are actually still exponentially localized
but on much larger scale then the states of the localized phase. Upon passing the 'mobility
edges' the inverse of the localization length changes from a quadratic to a quartic depen-
dence on the disorder strength, and for weak disorder it means that the increase of the spatial
extension of the electron wavefunctions can be huge. Therefore quantitatively this change
in the �nite length samples can be considered as a MIT. Similar conclusions have also been
reported for other disordered models with long-range correlations [83, 84].
18
Figure 2.4. The Anderson tight-binding model of a 1-D chain is depicted above.
"i 's correspond to on-site energies.
One more important study in the theory of localization for 1-D systems came from Hein-
richs [85, 86] who considered two-chain Anderson-model for a one-dimensional wire. In that
work, Heinrichs calculated analytically the inverse localization length for a disordered two-
chain wire with uncorrelated random potentials. The mentioned work has opened the way
to the calculation of localization length for two-channel wires with correlated disorder, which
is one of the major �ndings of this thesis and also the applied model for DNA electron
localization problems.
2.3. Analytical Results for Localization Length
Thouless has come up with an analytical solution of problem of Anderson localization
in a 1-D chain [17]. The tight-binding Anderson model of a chain (Fig.2.4) is represented
by a sequence of potential sites with on-site energies "n and constant amplitude of hopping
between the neighboring sites t. This is an example of the so-called diagonal disorder. The
quantum states in this chain are obtained from discrete Schrodinger equation
(2.10) t(�n+1 + �n�1) = (E � "n)�n:
The strength of the uctuations is measured by the variance "20 = h"2
ni. In case of weak
disorder, "0 � t, Thouless calculated the inverse localization length l�1(E) = 1l(E) in Born
approximation for white-noise potential, where h"i"ki = "20�ik . The Thouless formula for
inverse localization length has the following form
19
(2.11) l�10 (E) = (E) =
"20
8(1� E2=4t2):
Eq.(2.11) is the �rst term of the expansion of the Lyapunov exponent l�10 (E) over variance
"20. The validity of Born approximation suggests scattering is weak, there for that a localized
state covers many sites, i.e. l0(E)� 1. It becomes invalid in the vicinity of the band edges,
j E=2t j! 1, where l�10 ! 0. More accurate expansion over "0 in the vicinity of band edges
lead to unusual scaling, l�10 (E) / "2=3
0 [71, 72]. This "anomalous" behavior of the localization
length is later shown to be accompanied by violation of the scaling hypothesis [73, 74, 76].
Izrailev and Krokhin extended result (2.11) into a more general problem of Anderson-chain
with correlated disorder [18]. They showed that correlations in the random sequence of the
site energies "n strongly a�ect the interference pattern between the forward and backward
scattered wave. In the lowest Born approximation the localization length is determined by
a pair correlation function, h"i"ki = "20�(i � k), only. It was found that in the correlated
potential the localization length is modi�ed and it is given by the following formula
(2.12) l�10 (E) = (E) = '(2�); '(�) = 1 + 2
1∑
k=1
�(k) cos (�k) :
Here the parameter � plays the role of Bloch vector, de�ning the dispersion relation in a
perfectly periodic chain,
(2.13) E = 2t cos�:
It is clear from (2.12) that if '(�) vanishes, the inverse localization length also vanishes
and the localization length l(E) ! 1. This happens if the in�nite terms in Eq.(2.12)
contribute to the sum over k .Thus, the long-range correlations may lead to delocalized states
forming a band in the energy spectrum. This analytical expression as a novel result also
predicted the possibility of sharp mobility-edges.
20
Figure 2.5. Sharp mobility edge is a sudden change inverse localization length
moving along in the energy spectrum. The energy region l�1(E)= (E) = 0 is
the delocalized region. E = 1 line is the mobility edge.
Krokhin and Izrailev showed in their paper the necessary condition correlation function
shall satisfy to give a mobility-edge shown in Fig.2.5. The correlation function shall have the
form [18]
(2.14) �(k) =3
2�ksin(
2�k3
):
In 2000, Kuhl and St�ockmann , in collaboration with Izrailev and Krokhin have experimen-
tally demonstrated the existence of a mobility edge experimentally in a microwave transmission
experiment [87]. The setup they used (Fig.2.6) is a disordered microwave waveguide. They
took a toroidal waveguide, drilled holes on top of it with equal spacings in between. Screws of
same thickness were then placed at each hole. Each screw is a scatterer for the microwaves
and the depth of the screw determines the re ection coe�cient for that screw. It was the
random depths of screws that created the disorder in this 1-D like waveguide, and the depths
were chosen so that they generated a correlation function similar to Eq.(2.14).
The results of the experiment show (Fig.2.7) the mobility edge for a wave system. This
�nding has led me into believing that similar delocalized bands (or islands) in energy spec-
trum might also exist for DNA. My calculations on DNA molecules using Krokhin-Izrailev
localization length vs energy relation is given in Chapter 4.
At 2002, Heinrichs published his analytical expression for localization length in two-chain
Anderson-model of disordered wires. He generalized the tight-binding model Eq. (2.10) for
21
Figure 2.6. The photo and design of microwave transmission experiment setup
of Kuhl et. al. is shown above.
two-channel system (Fig.2.8) by introducing an index i at the wave function �n, and the
potential energy "n. This index denotes the channel, i = 1; 2. A term describing the inter-
channel coupling (with constant hopping parameter h) is added to the left-hand-side. The
second equation for the two-component wave function �i ;n is obtained by symmetrizing with
respect to the index i . Finally, Schr�odinger equation for the two-channel tight-binding model
is written as follows,
(2.15)t(�1;n+1 + �1;n�1) + h�2;n = (E � "1;n)�1;n ;
t(�2;n+1 + �2;n�1) + h�1;n = (E � "2;n)�2;n :
In his calculations, Heinrichs considered uncorrelated on-site energies, de�ning h"in"jmi =
"20�i j�nm. Then for the two-chain Anderson model, the inverse localization length (E) is
given by [85]
22
Figure 2.7. The transmissivity of the waveguide after a single pass is shown
below. The upper graph is transmission vs wavelength (energy is determined
by wavelength) for multiple passes. The straight line represents theoretical
assumptions, dashed line represents the experimental data.
Figure 2.8. The Anderson tight-binding model of a 2-D wire consists of two
connected chains.
(2.16) l�1(E) ="2
0
32
(1
sin�1+
1sin�2
)2
:
where similar to Eq.(2.13), the dispersion in the two-channels is given by,
(2.17)2 cos�1 = E � h;2 cos�2 = E + h :
.
23
The dispersion relations for wavenumbers �1 and �2 re ect the coupling between the two-
chains of the wire with coupling constant h creating a shift �h in the energy scale. Eq.(2.16)
is valid for energy intervals h�2 < E < 2�h, where both �1 and �2 are real and waves in both
channels are in propagating modes. Later Heinrichs generalized his solution to the intervals
�2� h < E < h�2 and 2� h < E < 2 + h, where one of the wave numbers �i is necessarily
a pure imaginary number and that wave is in evanescent mode decaying exponentially as a
function of the site number. The other mode is still propagating. In that case [86],
(2.18) l�1(E) ="2
0
161
sin�;
where � is the real wavenumber for the propagating mode.
In this chapter, I introduced the localization theory, its consequences and applications.
The last section summarized the analytical results available to us for calculation of local-
ization lengths in DNA. In the next chapter, the calculation of localization length for a
one-dimensional two-chain wire with correlated disorder is presented. The last chapter is
about the application of all analytical results for localization length to the speci�c problem of
electron localization in DNA.
24
CHAPTER 3
CALCULATION OF ELECTRON LOCALIZATION LENGTH FOR TWO-CHANNEL
WIRES
We seek to calculate electron localization length for two-channel disordered wires using
Eq.(2.9) given in Chapter 2. The inverse localization length l�1(E) is a functional of the
natural logarithm ln g of the dimensionless conductance g. To calculate conductance, we use
the Landauer two-probe conductance formula [88, 89]
(3.1) g = T r(t ty)
which describes current transport in a wire. Here t is the so-called transmission matrix of
the N-channel system
(3.2) t =
t11 t12 � � � t1N
t21 � � � � � � � � �� � � � � � � � �tN1 tN2 � � � tNN
:
An incoming wave from channel j of an ideal lead at one end of the disordered wire of
length L has a coe�cient ti j for transmission into channel i of the lead at the other end. The
Landauer conductance does not contain information about the contacts or the lead and it is
a sole property of the wire under consideration. We are interested in calculating the electron
localization properties in a stand-alone wire independent of the contacts, thus g as de�ned
above is suitable. The correctness of the Landauer formula has been experimentally validated
for perfectly transmitting channels [90, 91], the case when Eq.(3.2) reduces to
25
(3.3) g = N
where N is the total number of channels. Remark that the dimensionless conductance
g = G2e=h , where G is the conductance and 2e=h is the quantum of conductance.
We now proceed to the calculation of g. The tight-binding model of a one-dimensional
disordered two coupled-chain wire was introduced in Chapter 2. It was generated from a
single-electron Hamiltonian. The Hamiltonian is as follows,
(3.4) H =∑
i ;n
"i ;nji ; nihi ; nj+∑n
hj1; nih2; nj+ ∑
i ;n
tji ; nihi ; n + 1j+ h:c:
where site number n on chain i is represented by ji ; ni. Only the nearest-neighbor sites are
coupled with each other; the sites on the same chain are coupled through intra-chain coupling
parameter t and the sites of same number at di�erent chains are coupled through inter-chain
coupling parameter h. Electron-electron interaction is neglected in this model. The time-
independent Schr�odinger equation for the wire was given in Eq.(2.15). We will consider the
case h"i ;ni = 0, without loss of generality since it is possible to shift the E = 0 point on
the energy-axis to match the average onsite energy later. The randomness ( uctuations)
of the on-site energies is the origin of disorder in our model. The Schr�odinger equation
can be rewritten after normalizing all parameters with respect to t (h ! h=t,E ! E=t,
"i ;n ! "i ;n=t). We now write Eq.(2.15) normalized with respect to t, in matrix form,
(3.5)
�1;n+1 + �1;n�1
�2;n+1 + �2;n�1
=
E � "1;n �h�h E � "2;n
�1;n
�2;n
:
The value of �i ;n, amplitude of the wavefunction at site n on chain i , is coupled to the
values of �i ;n�1 and to the value of �j;n, the amplitude at the site n on the other chain.
The basis �i ;n do not represent uncoupled channels since even at the case all "i ;n ! 0 the
amplitudes of the wavefunction on chains 1 and 2 are still coupled, as shown below
26
(3.6)
�1;n+1 + �1;n�1
�2;n+1 + �2;n�1
=
E �h�h E
�1;n
�2;n
:
We want to use perturbation theory to calculate g. The second-order terms in pertur-
bation diverge if we work with a coupled basis. Thus we seek to diagonalize the interchain
coupling term h. The unitary transformation that leads to diagonalization is as follows,
(3.7)
1;n
2;n
=
1p2
1 1
1 �1
�1;n
�2;n
:
The Schr�odinger equation in the i ;n basis has the matrix form
(3.8) 1;n+1 + 2;n�1
2;n+1 + 2;n�1
=
E � h � 1
2 ("1;n + "2;n) 12 ("2;n � "1;n)
12 ("2;n � "1;n) E + h � 1
2 ("1;n + "2;n)
1;n
2;n
:
Now Schr�odinger equation is written in the "right" basis. In this basis, the wire is a sys-
tem of two uncoupled channels for electron wave transmission. In what follows, we develop
a perturbation theory for this equation.
3.1. Transfer-Matrix Formalism
The matrix equation (3.8) relates the amplitudes of wavefunction i ;n at site n to the
amplitudes at the nearest neighbor sites. Now we focus the attention at freely propagat-
ing waves at the leads. The Landauer conductance formula gives conductance in terms of
transmission coe�cients ti j for propagating waves in leads through the wire. The leads have
a perfectly periodic structure and the wire have energy uctuations on periodically arranged
sites (there is no topological disorder). The propagating waves on the leads are nothing but
Bloch waves (2.3) satisfying
27
(3.9) �i ;n+1 = e�{� �i ;n
de�ning +i ;n ( �i ;n) as amplitudes of a wave propagating from left-to-right (right-to-left)
in channel i .
In the transfer-matrix formalism, the propagation of a wave from left-to-right is repre-
sented as a matrix equation,
(3.10)
'j+1
'j
= Xj
'j
'j�1
where Xj is the so-called transfer-matrix for site j . Matrix Xj translates the wave from
the left of site j to the right of site j . For our two-channel wire, Eq.(3.10) is replaced by a
4-dimensional matrix equation,
(3.11)
1;n+1
1;n
2;n+1
2;n
= ~Xn
1;n
1;n�1
2;n
2;n�1
The transfer matrix ~Xn can be derived from (3.8) according to the de�nition given above,
yielding
(3.12)
1;n+1
1;n
2;n+1
2;n
=
E � h � an �1 bn 0
1 0 0 0
bn 0 E + h + an �1
0 0 1 0
︸ ︷︷ ︸~Xn
1;n
1;n�1
2;n
2;n�1
where an = 12 ("1;n + "2;n) and bn = 1
2 ("2;n � "1;n).
28
The sites located at the ideal leads have no uctuations in energy (no disorder) with all
on-site energies "i ;n ! 0 and an = bn = 0 as well. As discussed before, the solutions of waves
propagating along the leads are Bloch waves in form (3.9). If we insert +i ( �i ) into (3.12),
we get for the leads
(3.13)
e�{� 1;n
e�{� 1;n�1
e�{� 2;n
e�{� 2;n�1
=
E � h �1 0 0
1 0 0 0
0 0 E + h �1
0 0 1 0
1;n
1;n�1
2;n
2;n�1
:
We can now calculate the wavevector � for the Bloch wave from by simply solving the
eigenvalue problem (3.13), ~Xn +i = e {� +
i ( ~Xn �i = e�{� �i ). After simple algebra, the
dispersion relations for the waves propagating in both channels take the form
(3.14)2 cos�1 = E � h;2 cos�2 = E + h :
.
The eigenfunctions for the leads are of the form of plane waves,
(3.15) �i ;n = e�{n�i ;
where �i are taken to be positive, 0 � �i � � so that these functions correspond to plane
waves traveling from left to right and from right to left, respectively.
The transfer matrices for the single slices n (Fig.3.1) are diagonalized in the basis of the
plane waves +i ;n ( �i ;n) that satisfy the Bloch conditions of the system. The transfer matrix
Xnfor slice n in that basis is,
29
Figure 3.1. Sites n on both chains are taken as a slice in the channel basis
i ;n. The transfer matrix Xn corresponding to slice n translates the waves by
a single site, adding phases or scattering the waves traveling in both directions.
(3.16)
Xn = X0+{Xn =
e {�1 0 0 0
0 e�{�1 0 0
0 0 e {�2 0
0 0 0 e�{�2
+{
�1ne {�1 �1ne�{�1 ��ne {�2 ��ne�{�2
��1ne {�1 ��1ne�{�1 �ne {�2 �ne�{�2
��ne {�1 ��ne�{�1 �2ne {�2 �2ne�{�2
�ne {�1 �ne�{�1 ��2ne {�2 ��2ne�{�2
;
where X0 is the transfer matrix of the periodic chain and X(1)n is the correction due to
weak disorder. X(1)n is responsible for scattering processes that take place while the waves
travel along the channels. The parameters �in and �n are de�ned as
(3.17) �1n ="1;n + "2;n
4 sin�1; �2n =
"1;n + "2;n
4 sin�2; �n =
"2;n � "1;n
4p
sin�1 sin�2:
The role of the transfer matrix Xn is to relate the amplitudes of the plane waves on the
left and right side of the slice,
(3.18)
+1;n+1
�1;n+1
+2;n+1
�2;n+1
= Xn
+1;n
�1;n
+2;n
�2;n
:
30
Now we have the transfer matrix for a single slice and the transfer matrix X corresponding
to the whole system is the product
(3.19) X =L∏n=1
Xn
The total length of the system is La where a is the distance between each slice. The
transfer matrix of the whole systems relates the amplitude of the waves at both ends of the
wire, similar to (3.18),
(3.20)
+1;L
�1;L +
2;L
�2;L
= X
+1;0
�1;0
+2;0
�2;0
:
3.2. Lyapunov Exponent
The matrices Xn are random matrices, because the part X(1)n contains random on-site
energies "1;n and "2;n of the disordered wire. Since we are considering in�nite length wires
where L ! 1, the �nal transfer matrix is a multiplication of in�nite number of random
matrices.
Lyapunov exponents emerge from the random matrix theory [92], and they are used to
characterize the asymptotic behavior of the systems determined by the products of such
matrices. The ergodic theorem of Oseledec [93] and Tutubalin [94] states that for a product
of in�nite number of random matrices belonging to the same statistical ensemble
(3.21) ML = MLML�1 � � �M1;
there exists a limiting matrix �,
(3.22) limL!1(Mt
LML)1
2L � � � 0;
31
and its eigenvalues can be written as e�i . The maximum of these Lyapunov characteristic
exponents = max(�1; �2; � � �; �d) is given by
(3.23) = limL!1
1L
ln(ML):
The Lyapunov exponent (3.23) is directly related to Eq.(2.9) for the inverse localization
length. The relation of dimensionless conductance g to the product of random transfer
matrices Xn is shown in the following section in detail. It is su�cient at the moment to
say that in the limit L ! 1, Oseledec's theorem dictates that we get the maximum of the
Lyapunov coe�cients. This Lyapunov exponent, in relation to the formula in Eq.(2.9), gives
the minimum localization length. The minimum localization length is the one that determines
the mobility of the electrons with a given energy inside a disordered medium.
3.3. Scattering Matrix and Conductance
The scattering of plane waves -re ection and transmission- at and between the two ends
of the 1-D systems is governed by the scattering matrix S ,
(3.24) S =
r�+ t��
t++ r+�
;
where
(3.25) t�� =
t��11 t��12
t��21 t��22
; r�� =
r��11 r��12
r��21 r��22
:
Here t++i j (t��i j ) and r�+
i j (r+�i j ) are the transmission and re ection amplitudes in the channel
i , provided there is a unit ux incident from left(right) in the channel j . The transmission
matrix t in Landauer conductance formula (3.1) is nothing but one of the transfer matrices
t�� or t++, which are equal since the system has time-reversal invariance.
In general, the scattering matrix expresses outgoing wave amplitudes in terms of ingoing
ones on either side of the quasi-1D disordered wire via the scattering relations [95, 96],
32
(3.26)
O
O0
= S
I
I 0
:
Here I and I 0 (O and O0) denote ingoing (outgoing) amplitudes at the left and right sides
of the disordered region, respectively.
In our case, the matrix S is de�ned as follows
(3.27)
�1;0
�2;0
+1;L
+2;L
= S
+1;0
+2;0
�1;L �2;L
:
Since
(3.28)
+1;L
�1;L +
2;L
�2;L
=
X11 X12 X13 X14
X21 X22 X23 X24
X31 X32 X33 X34
X41 X42 X43 X44
+1;0
�1;0
+2;0
�2;0
;
it follows from simple algebra that
(3.29) t = t�� =1�
X44 �X24
�X42 X22
;
(3.30) � = X22X44 �X24X42:
Finally, the dimensionless conductance g can be written in terms of elements of transmis-
sion matrix X,
(3.31) g =�j�j2 =
jX22j2 + jX44j2 + jX42j2 + jX24j2jX22X44 �X24X42j2 :
33
3.4. Calculation of Transfer Matrix in Second-Order Born Approximation
The total scattering matrix X is a product matrices represented with a diagonal part X0
plus non-diagonal part X(1)n that depends linearly on the on-site energies.
(3.32)
X =∏Ln=1(X0 +X(1)
n = XL0 + {
∑n X
L�n0 � X(1)
n �Xn�10 �∑
m>n XL�m�n0 �X(1)
m � Xm�n�10 � X(1)
n �Xn�10
+F (O("3)) + G(O("4)) + � � �
where F (O("3)) (G(O("4))) is a function of the parameter in third-order (fourth-order) in
", the disorder in on-site energies. We limit our discussion to the weak disorder case, where
(3.33) "i ;n << 1
and continue our calculations within second-order in Born approximation. So, the total
transfer matrix X will only include terms that are smaller or equal to second-order in ". The
�rst three-terms in (3.32) remain in X = A+ {B � C, where
(3.34) A = XL0 =
e {�1L 0 0 0
0 e�{�1L 0 0
0 0 e {�2L 0
0 0 0 e�{�2L
;
(3.35)
B =∑n
e {�1L�1n e {�1Le�2{�1n�1n �e {�1Le {(�2��1)n�n �e {�1Le�{(�1+�2)n�n
�e�{�1Le2{�1n�1n �e�{�1L�1n e�{�1Le�{(�1+�2)n�n e�{�1Le {(�1��2)n�n
�e {�2Le {(�1��2)n�n �e {�2Le�{(�1+�2)n�n e {�2L�2n e {�2Le�2{�2n�2n
e�{�2Le {(�1+�2)n�n e�{�2Le�{(�1��2)n�n �e�{�2Le2{�2n�2n �e�{�2L�2n
;
and the terms in C that we need (C22,C24,C42 and C44),
34
C22 =∑m>n
e�{�1L[(1� e2{�1(m�n))�1m�1n+(e {(�1��2)(m�n) � e {(�1+�2)(m�n))�m�n];(3.36)
C44 =∑m>n
e�{�2L[(1� e2{�2(m�n))�2m�2n+(e {(�2��1)(m�n) � e {(�1+�2)(m�n))�m�n];(3.37)
C24 =∑m>n
e�{�1Le {(m�1�n�2)[(e {�2(m�n) � e�{�2(m�n))�m�2n+(e {�1(m�n) � e�{�1(m�n))�m�1n];
(3.38)
C42 =∑m>n
e�{�2Le {(m�2�n�1)[(e {�1(m�n) � e�{�1(m�n))�m�1n+(e {�2(m�n) � e�{�2(m�n))�m�2n]:
(3.39)
Given below are the elements of the total transmission matrix X that we will use,
(3.40)
X22 = e�i�1L[1 + i
∑m �1m �∑
m>n(1� e2i�1(m�n))�1m�1n
�∑m>n(e i(�1��2)(m�n) � e i(�1+�2)(m�n))�m�n
];
X44 = e�i�2L[1 + i
∑m �2m �∑
m>n(1� e2i�2(m�n))�2m�2n
�∑m>n(e i(�2��1)(m�n) � e i(�1+�2)(m�n))�m�n
];
X24 = �e�i�1L[i∑
m ei(�1��2)m�m +O("2)];
X42 = �e�i�2L[i∑
m ei(�2��1)m�m +O("2)]:
The term � in Eq.(3.31)is
(3.41)
� = e�{(�1+�2)L[1 + {∑n
(�1n + �2n)
�∑m>n
(�1m�2n � e {(�1��2)(m�n)�m�n)
�∑m>n
[e {(�1��2)(m�n) + e�{(�1��2)(m�n) � 2e {(�1+�2)(m�n)]�m�n
�∑m>n
[(1� e2{�1(m�n))�1m�1n + (1� e2{�2(m�n))�2m�2n]];
with
35
(3.42)
j�j2 = 1+∑m;n
(�1m + �2m)(�1n + �2n)� 2∑m;n
(�1m�2n) + 2∑m;n
cos(k1 � k2)(m � n)�m�n
� 4∑m>n
cos(�1 � �2)(m � n)�m�n � 4∑m>n
cos(�1 + �2)(m � n)�m�n
� 2∑m>n
[(1� cos 2�1(m � n))�1m�1n + (1� cos 2�2(m � n))�2m�2n]:
We can simplify the equality a bit more by noting
(3.43)∑m;n
f (m; n) =∑n
f (n; n) + 2∑m>n
f (m; n)
so that
(3.44)
j�j2 =1 +∑n
[�21n + �2
2n + 2�2n ]
� 4∑m>n
cos(�1 + �2)(m � n)�m�n
+ 2∑m>n
[cos 2�1(m � n)�1m�2n + cos 2�2(m � n)�2m�1n]:
Similar calculations yield � = 1 + j�j2. At the end, we get for inverse localization length
l�1
(3.45) l�1 = � limL!1
12Lhln(
1 + j�j2j�j2 )i:
Let's take a look at j�j2;
(3.46) j�j2 = 1 + �("2)) ln(1 + j�j2j�j2 ) ' ��
2
when we keep only the second-order terms in ". This brings us to,
36
(3.47)
l�1 = limL!1
14Lh∑m
�21m + �2
2m + 2�2mi
+ 4h∑p
cos((�1 + �2)p)∑n
�n�n+pi
+ 2h∑p
[cos(2�1p)∑n
�1n�1(n+p) + cos(2�2p)∑n
�2n�2(n+p)]i:
The result is in terms of �'s and �'s, previously de�ned in (3.17). If we trace the
conversions back,
(3.48)
�21m =
("1;m + "2;m)2
16 sin2 �1="2
1;m + "22;m + 2"1;m"2;m
16 sin2 �1;
�22m =
"21;m + "2
2;m + 2"1;m"2;m
16 sin2 �2;
�2m =
"21;m + "2
2;m � 2"1;m"2;m
16 sin�1 sin�2;
and also,
(3.49)�1m�1(m+p) =
"1;m"1;(m+p) + "2;m"2;(m+p) + "2;m"1;(m+p) + "1m"2;(m+p)
16 sin2 �1;
�m�m+p ="1;m"1;(m+p) + "2;m"2;(m+p) � "2;m"1;(m+p) � "1;m"2;(m+p)
16 sin�1 sin�2:
De�ning correlation functions �i j(p) as follows
(3.50)
"210�11(p) =
∑m "1m"1(m+p)
L;
"220�22(p) =
∑m "2m"2(m+p)
L;
"12�12(p) =∑
m "1m"2(m+p)
L;
we obtain the �nal result for the inverse localization length l�1(E) as given below :
37
l�1(E) ="2
10
64
['11(2�1)
sin2 �1+'11(2�2)
sin2 �2+
2'11(�1 + �2)sin�1 sin�2
]
+"2
20
64
['22(2�2)
sin2 �2+'22(2�1)
sin2 �1+
2'22(�1 + �2)sin�1 sin�2
]
+"12
32
['12(2�1)
sin2 �1+'12(2�2)
sin2 �2� 2'12(�1 + �2)
sin�1 sin�2
]:(3.51)
The structure of this formula for the inverse localization length is similar to the corresponding
formula (2.12) for a single-channel system. The localization of an electron occurs due to
elastic backscattering processes in both channels with change of the momentum by 2�1
and 2�2 and due to inter-channel scattering with change of the momentum by �1 + �2.
An unperturbed wave, which according to Eq.(3.7) is either symmetric or anti-symmetric is
scattered at three random potentials with variances "21, "2
2, and "12. The correlations enter
through the functions 'i j(�), which are represented by the Fourier series,
(3.52) 'i j(�) = 1 + 21∑
k=1
�i j(k) cos(�k):
The inverse localization length (3.51) is positively de�ned for all the allowed energies.
In the case of uncoupled channels, h = 0, the dispersion relations (3.14) are identical,
�1 = �2 = �, and the cross-correlation term vanishes. As a result Eq.(3.51) is simpli�ed to
the following form
(3.53) l�1(E) =1
sin2 �["2
1'11(2�) + "22'22(2�)
]:
For identical channels, "10 = "20 = "0 and '11(2�) = '22(2�) the single-chain case (2.12) is
recovered. Finally, if the potentials in the both channels are uncorrelated (white-noise) then
'i j(�) � 1 and "12 = 0. In this case (3.51) reproduces Heinrichs' result (2.16) [85],
(3.54) l�1(E) ="2
0
32
(1
sin�1+
1sin�2
)2
:
3.5. One Evanescent and One Propagating Mode
In the energy regions �2 � h < E < h � 2 and 2 � h < E < 2 + h, one of the wave
numbers �i is necessarily a pure imaginary number. The corresponding wave function decays
38
exponentially away from the entering point with decrement j �i j. Since the transmission
matrix is a relation between the propagating wave modes only, the evanescent modes do not
contribute to the conductance of a long sample if L >>j �i j�1.
It was shown in Heinrichs' paper [85] that in the case of uncorrelated disorder the evanes-
cent mode term does not contribute to the Lyapunov exponent (2.16) and has to be omitted.
Moreover, the coupling between the propagating and evanescent modes is strongly suppressed.
This results in an extra factor of 2 in (2.16). In what follows I demonstrate that this sce-
nario of transition from propagating to evanescent regime remains unchanged in the case of
correlated potential.
It is worthwhile to note that in weak disorder approximation the formulas for the Lya-
punov exponents are invalid in the vicinity of the critical energies Ec = �2 � h, where the
transition from propagating to evanescent regime occurs. At these energies the perturbation
"i= sin�i !1 and the Born approximation fails even for weak disorder.
Let us consider the energy domain 2�h < E < 2+h where the second mode is evanescent,
�2 = i�. The transfer matrix of the nth site, Xn, establishes a linear relation between the
wave functions on the both side of this site. Since the evanescent mode does not contribute
to the conductance, we are interested only on the elements of the transfer matrix which
describe scattering from propagating to propagating mode. For the propagating mode with
index 1 those elements are at the 2x2 upper left block of the 4x4 transmission matrix Xn,
(3.55)
+1;n+1
�1;n+1
...
=
Xn;11 Xn;12 : : :
Xn;21 Xn;22 : : :...
......
+1;n+1
�1;n+1
...
:
Now we can introduce a transfer matrix for the nth site, using the basis of the propagating
modes only
(3.56)
+
1;n+1
�1;n+1
= Tn
+
1;n
�1;n
; Tn =
Xn;11 Xn;12
Xn;21 Xn;22
:
39
Similar to the transfer matrix Xn , the transfer matrix Tn has a zero-order diagonal component
and linear over the random potential term,
(3.57) Tn = T0 + T (1)n =
e i�1 0
0 e�i�1
+
ie i�1an ie�i�1
�ie i�1an �ie�i�1
:
The matrix Tn, being a unitary matrix, conserves the ux. The total transfer matrix T of
a sample is a product of all Tn matrices. Keeping up to quadratic over disorder terms in this
product, a formula similar to Eq.(3.32) is obtained,
(3.58) T =L∏n=1
Tn = T L0 +L∑m
T L�m0 � T (1)m � Tm�1
0 +∑m>n
T L�m0 � T (1)m � Tm�n�1
0 � T (1)n � T n�1
0 :
In the presence of evanescent mode the dimension of the scattering matrix is also reduced
since now it relates the incoming and outgoing components of the propagating wave only,
(3.59)
+
1;0
�1;0
= T
+
1;L
�1;L
;
�1;0
+1;L
= S
+
1;0
�1;L
:
The transmission matrix t, which determines the conductance becomes a scalar. It is obtained
from the scattering matrix,
(3.60) S = 1T11
T21 1
1 �T12
; t ty =j T11 j�2 :
Thus, in the presence of evanescent mode the conductance (and hence the localization
length) is determined by j T11 j2. This quantity can be calculated from (3.58) and in the
quadratic approximation we get,
(3.61) jT11j2 = 1 +∑m;n
�m�n � 2∑m>n
�m�n + 2∑m>n
cos 2�1(m � n)�m�n:
Now, expanding ln t ty = � ln j T11 j2, the following result for the inverse localization
length is obtained,
(3.62) l�1(E) =1
32 sin2 �1
["2
1'11(2�1) + "22'22(2�1) + 2"12'12(2�1)
]:
40
In the counterpart region �2� h < E < �2 + h the �rst mode becomes evanescent and the
Bloch number �1 in (3.62) is replaced by �2.
The localization length is a complicated linear functional of the correlation functions
�ik(k). Analysis of the localization length for di�erent classes of short- and long-range cor-
relations requires more work. It is not clear yet what kind of long-range correlations are
su�cient for the mobility edge to appear in the spectrum of two-channel system. Here I
give numerical results for uncorrelated potential and for exponentially decaying correlation
function of the following form
(3.63) �i j(k) = (�1)ke��jkj
where � is the inverse radius of correlations. Here correlations alternate with anti-correlations.
The localization length for this speci�c correlation function shows oscillatory behavior in the
interval of E where both channels are open. Unlike this, in the regions where only one of
the channels is propagating the localization length is a monotonic function of energy. There
is a discontinuous jump for l(E) at the critical energies E = �(2 � h) where a transition
from propagating to evanescent mode occurs. These discontinuities are clear evidences of
the fact that Born approximation is not valid in the vicinities of the critical points. Extended
states do not appear for this class of short-range correlations. They probably may appear if
the correlations are of long-range, i.e. the correlations decay as a power law.
3.6. Discussions
In the single-channel case the "inverse" scattering problem - reconstruction of the statis-
tical ensemble of the correlated potentials through the localization length l(E) - has a unique
solution. It follows from the general properties of the Fourier integrals that if the Lyapunov
exponent vanishes within a �nite interval of energies, the correlation function decays as a
power law, �(k) / 1=kp. For a sharp mobility edge the parameter p = 1 [75, 18, 76].
In a two-channel system the inverse scattering problem is more complicated since there
are three correlators, �11(k), �22(k), and �12(k), which have to be reconstructed from the
function l�1(E), given by Eqs.(3.51) and (3.62). However, it is clear from the structure of
41
0
50
100
150
200
250
-2 -1 0 1 2
l(E
)
Energy (eV)
Figure 3.2. The localization length vs energy graphs for various correlated/non-
correlated disorder is given above. Solid line is for uncorrelated (white-noise)
potential. Dashed line is for exponential intra- and inter-channel correlation
functions, all three are given by (3.63). Dotted line is for exponential intra-
channel correlations and delta-correlated inter-channel scattering, h"i"ji / �i j .The parameters of the model are h = 0:5 eV, t = 1:0 eV and h"2
1i = h"22i =
h"212i = 0:252.
Eqs. (2.12), (3.51) and (3.62) that the power-decaying correlation functions are necessary
in order to have a mobility edge. If the correlations are of a short-range, only a discrete set
of resonant extended states may appear in the spectrum.
In particular, an extended state at the band center E = 0 was predicted for a two-channel
random dimer for some speci�c parameters of the random potential [77]. The random dimer
is an example of a dichotomous sequence where the on-site potential takes on only two
values, e.g, "0 and �"0. For all sites "1;n = "2;n. At a site n the value "0 or �"0 emerges
with probability of 1/2 and it is repeated at the nearest site n + 1. Statistical properties of
this model are characterized by the mean value h"ni = 0, variances⟨"2
1;n
⟩=
⟨"2
2;n
⟩= "2
0,
"12 = h"1;n"2;ni = "20 and the correlators,
42
(3.64) �11(1) = �22(1) = �12(1) = h"n"n�1i = 1=2:
and �i j(k > 1) = 0. By substituting this correlator into Eq. (3.51), we obtain the inverse
localization length for the dimer,
(3.65) l�1(E) ="2
0
8(
cot2 �1 + cot2 �2):
This function does not vanish, if the inter-channel hopping parameter h is di�erent from
zero, i.e. there is no extended state in a two-channel dimer. Unlike this, in a single-channel
dimer there are two extended states, which in the case of weak disorder are situated in the
vicinity of the band center E = 0 [55]. Indeed, the Lyapunov exponent Eq.(3.65) vanishes
quadratically at a single point E = 0 if �1 = �2 = �=2. Eq.(3.65) is valid for weakly
disordered potential, when "0 � 1, therefore it does not show the existence of the extended
state at E = 0 for the dimer with "0 = 1 [77].
Finally, I applied the obtained results to the two-stranded DNA molecule to make an
analytical investigation. Caetano and Schultz used a DNA-like structure where the four basic
nucleotides are evenly represented in a molecule and both strands form uncorrelated sequences
of the nucleotides [78]. In terms of my parameters it means that "12 = ("A"T + "C"G)=2,⟨"2
1;n
⟩=
⟨"2
2;n
⟩= ("2
A + "2T + "2
C + "2G) =4, and h"1;n+k"1;ni = h"2;n+k"2;ni = 0. Substituting
theses values into Eq.(3.51) we get,
(3.66)l�1(E) = 1
128
[("A + "T )2 + ("C + "G)2] (
1sin2 �1
+ 1sin2 �2
)+
164 sin�1 sin�2
[("A � "T )2 + ("C � "G)2] :
Here the electron energy E, which enters through the dispersion relations Eq.(4.5) and the
on-site energies of the nucleotides are counted from the mean site energy h"1;ni. Since sin�1
and sin�2 are positive functions (0 � �1; �2 � �), the Lyapunov exponent Eq.(3.66) is a sum
of two positive quantities, independently on the particular values of the on-site energies. This
43
means that the discussed model of DNA does not allow existence of the extended states.
A band of extended states were predicted in Ref.[78], using numerical simulations of the
inverse participation ratio. Later this result has been criticized in Ref.[79] on the basis of
group theory arguments. Formula (3.66) explicitly shows that in a two-channel system the
extended states cannot appear solely due to base pairing A-T, C-G. The base pairing indeed
decreases the Lyapunov exponent Eq.(3.51) because of the negative value of the parameter
"12, but this pairing (even being strong) is not enough to give rise to the extended states in
the uncorrelated DNA strands. Unlike this, the long-range correlations in a single-stranded
model of DNA may lead to a band of extended states [99].
44
CHAPTER 4
ELECTRON LOCALIZATION IN DNA
Electrons in a DNA molecule can be classi�ed in two groups : The �rst group consists
of the electrons that are strongly bound to the molecular structures of the nucleic bases and
sugar phosphate groups. These electrons' wavefunctions are restricted to the spatial extent
of the host molecules. The second group of electrons are those that are loosely bound to the
molecules, somewhat free to move (or hop) from one molecule to another, i.e. from base to
base or base to sugar phosphate groups. Unlike the �rst group of electrons, the second group
of electrons have wavefunctions that cover several base lengths. They are also the electrons
that contribute to electronic transport in DNA. We speak of the localization of that second
group of electrons.
The structure of DNA was discussed in Section 1.1. As mentioned there, the most
likely mechanism responsible for the electron transport is the hopping between bases due
to the overlap of the �-orbitals (� stacking) [97]. Electrons occupying these orbitals have
�nite quantum mechanical probabilities to jump to the neighboring molecular orbital. The
interactions of electrons with the rest of the system is taken into consideration when molecular
orbital energies and hopping parameters are calculated. This fact makes it possible to develop
a single-electron Hamiltonian when solving for the eigenfunctions and transport properties of
the whole DNA molecule.
Two popular models for electronic transport in DNA are considered in this thesis, the
Fishbone model and the two-channel wire model. Both are based on tight-binding single-
electron Hamiltonians. All calculations are carried out at T = 0 K so coupling to phonons
etc. is thoroughly ignored. In both models, parameters for electron hopping and on-site
energies are borrowed from ab-initio calculations that strongly agree with empirical �ndings
[37, 98].
45
4.1. Fishbone Model of DNA
In the Fishbone model of DNA, electron transport in a DNA molecule is considered as
discrete jumps between the neighboring base nucleotides. Due to strong A-T and C-G cou-
pling, it is usually considered that there is a single electron channel with a binary sequence
of the on-site potentials with energies "A�T = 8:69 eV and "G�C = 8:31 eV. These on-site
energies are taken as the average of the ionization energies of the bases that form the base
pair. The corresponding ionization energies are "A = 8:24 eV, "T = 9:14 eV, "C = 8:87 eV
and "G = 7:75 eV. Each site (base pairs A-T and C-G) in this sequence has a link to the
sugar molecules, see Fig.4.1. The sugar molecules attached to the phosphate groups form
the backbone structure. From the base site an electron may jump either to the neighboring
sites or to one of the backbone sugars. The backbone sugars are not coupled to another
backbone sugar. Fig.(4.1) represents the main features of this so-called �shbone model.
The Hamiltonian of the �shbone model is written as follows [98, 99]:
(4.1) HF = �L∑
i=1
ti jiihi + 1j �L∑
i=1;q=u;d
tqi jqiihi j+∑
i ;qi
"i jiihi j+ "qi jqiihqi j+ h:c: :
Here jii and jqii are the wave functions at the base and backbone site respectively, ti is
the hopping parameter between base sites i and i + 1, tqi is the hopping between the upper,
q = u (lower, q = l)) backbone site and the base site, and "i and "qi are the on-site energies
of the bases and the backbones, respectively. In what follows I will consider the case of the
diagonal disorder when the hopping parameters ti = t and tui = tdi = tb are coordinate-
independent constants. The role of the o�-diagonal disorder was considered in Zhang and
Ulloa's paper [100]. It is convenient to chose t as a unit of energy, so as all the energies
become dimensionless. The randomness in the structure of the DNA molecules is due to
the weak uctuations of the on-site energies, "i = "0 + �"i and "u;di = "b + �"u;di , where
h�"ii2; h�"u;di i2 � 1. After these simpli�cations the Schrodinger equation corresponding to
Hamiltonian (4.1) can be written in the tight-binding model, similar to Eq.(2.10),
(4.2) i+1 + i�1 = (�i � E) i :
46
Figure 4.1. The depiction Fishbone model for DNA is shown above. The base
pairs and backbones have disordered on-site energies while the hopping param-
eters t and tb are constant. The base pairs are either A-T or G-C, with energies
"A�T = 8:69 eV and "G�C = 8:31 eV respectively.
where in place of the on-site energy �i , we put �i [99],
(4.3) �i = �"i +t2b
(E � "b)2f�"ui + �"di g;
and the eigenenergy E is replaced by
(4.4) ~E = E � "0 � 2t2b
E � "b :
The e�ective energy ~E, being the eigenenergy of the tight-binding model Eq.(4.2), lies within
the interval �2 < ~E < 2. The dispersion relation for ~E has a standard form, ~E = 2 cos�.
The dispersion relation for the electron energy E is obtained from Eq.(4.4),
(4.5) E�(�) ="0 + "b
2+ cos�� 1
2
√("b + "0 + 2 cos�)2 � 4
["b("0 + 2 cos�)� 2t2
b
]:
In the �shbone model the allowed energies form two equivalent conduction bands [98], thus
modeling semiconductor behavior of DNA observed in the experiment [13].
47
8
8.5
9
9.5
10
10.5
0 0.5 1 1.5 2 2.5 3
"enermin_vs_mu""enerplu_vs_mu"
8.5
Figure 4.2. The E� dispersion curves for the �shbone model are shown above
as functions of �.
Eq.(4.3) is a result of linear expansion over weak uctuations, therefore both terms are
supposed to be small. Because of the singularity in the second term, a narrow interval of
energies close to E = "b is excluded from the consideration.
The genetic information is coded in the sequence of potentials "i , therefore the uctua-
tions �"i are correlated. Unlike this, the uctuations of the potentials in the two backbone
sequences do not carry genetic information, therefore they are considered to be uncorrelated
and statistically independent. Thus, the correlations that exist in the sequence of the random
numbers �i are due to the correlations between the base pairs only,
(4.6) h�i�ki = �20�(i � k) + �2 t4
b
(E � "b)2 �ik :
Here �20 = h�"2
i i, �2 = h(�"ui )2i + h(�"di )2i, and �(k) is the normalized (�(0) = 1) binary
correlation function of the base sequence.
After the calculation of the binary correlation functions, I employed the �ndings of Izrailev
and Krokhin [18] that relates the long-range binary correlation function to the inverse lo-
calization length, given in Eq.(2.12). In the next section, I will summarize the numerical
calculations on localization lengths of some real human DNA molecules.
48
-0.2
0
0.2
0.4
0.6
0.8
1
0 200 400 600 800 1000
Cor
rela
tion
Fun
ctio
n
k
BRCA
0
0.5
1
0 2 4 6 8 10k
-0.02
0
0.02
0.04
100 110 120k
BRCA
-0.02
0
0.02
0.04
100 110 120k
BRCA
Figure 4.3. Binary correlation function of the sequence of nucleotides of Hu-
man BRCA gene exon region 11 is given above. The correlation function drops
from the value of 1 at k = 0 to � � 0:1 at k � 1, left insert. Correlations
extend to distances of a few thousands of base pairs, decaying very slowly.
An important feature of this correlation function is close to regular oscillations
about zero, right insert.
4.2. Numerical Calculations
The numerical calculations in this thesis are based on the information about DNA se-
quences available from the GenBank database [101]. I analyzed a number of DNA's and
found a few of them possessing a mobility edge.
To calculate the correlation function �(k), a sequence of nucleotides was mapped into
a binary one with energies �("A�T � "G�C)=2t. The value of t = 0:37eV [99] and the
average backbone energy is taken equal to �0 since the backbones and the nucleotides in the
base sequence are chemically connected. The correlation functions for two di�erent DNA
segments are shown in Figs. 4.3 and 4.4. Both correlation functions oscillates with small
but slowly decaying amplitude. These oscillations turn out to be the source of the sharp
49
-0.2
0
0.2
0.4
0.6
0.8
1
0 200 400 600 800 1000
Cor
rela
tion
Fun
ctio
n
k
BRCA
0
0.5
1
0 2 4 6 8 10k
-0.02
0
0.02
0.04
100 110 120k
BRCA
-0.02
0
0.02
0.04
100 110 120k
BRCA
Figure 4.4. Binary correlation function of the sequence of nucleotides of 51
kbp (kilo base pairs) region of Human Chromosome 22 is graphed above. The
region includes both exon and intron regions. The correlation function exhibits
behavior similar to that shown in Fig. 4.3 but with oscillations about a small
positive value.
metal-insulator transition predicted in Izrailev and Krokhin's paper [18] for the correlation
function Eq.(2.14), which oscillates with power-decaying amplitude.
In the numerical calculations the value of k in Eq.(4.6) does not run to in�nity but is cut
at the value of kmax = 400. This provides good convergence of the Fourier series and an
accuracy of < 1% for the localization length.
The localization length, corresponding to the correlation functions in �gures Fig.4.3 and
Fig.4.4 is plotted in �gures Fig.4.5 and Fig.4.6. Only the region of high-energy band (see
Fig.4.2) is shown, since the dependence l(E) in the low-energy band is exactly the same. In
the both cases the major part of the allowed energies is occupied by the states localized at
the distances of 30-70 base pairs. At these energies DNA macromolecule of the length of
� 103 behaves like an insulator. However, there are sharp peaks near E = 10:2 eV. At the
peak regions the localization length is orders of magnitude larger. My estimates showed that
50
Figure 4.5. Localization length vs Energy for Human BRCA gene exon region
11 was calculated calculated using Fishbone model.
it can be as long as 104 � 105 base pairs. In the experiments usually much shorter segments
are examined, which, thus, may exhibit metallic behavior within the peak regions of energies.
In Fig.4.5 and Fig.4.6 the peak regions are shown in the inserts. However, the localization
length there remains �nite, very sharp increase of the localization length should be considered
as a mobility edge. Within the peak region the states with localization length l(E) � 105
are practically delocalized. The region of the delocalized sates occupies a narrow interval of
energies � 20 meV.
For the both DNA's, the presence of the sharp mobility edges is due to the oscillatory be-
havior of the correlation function. The long-range correlations by themselves do not necessary
lead to the metal-insulator transition. The sharp, step-like transition requires the correlation
function given by Eq.(2.14). Any deviation from this speci�c form leads to the "softening"
of the mobility edge and to its disappearance.
4.3. Two-Chain Model of DNA
The Fishbone model of DNA considers only a single-channel for electron propagation.
Electron hops from one base pair to another. A better approximation requires taking into
account the details of the base pairs. The two bases forming a pair have di�erent ionization
51
Figure 4.6. Localization length vs Energy for a part of Human Chromosome
22 DNA graphed above was calculated using Fishbone model.
energies and contribute to electron hopping separately. The electron may jump from one
base to the other base in the base pair or to a neighboring base on the same strand. The
statistics of the chain changes with respect to Fishbone model, because we have four distinct
values of on-site energies instead of a binary one.
We model the DNA molecule as a one-dimensional two-chain disordered wire, in exact
accord with the wires explained in Chapter 3. The strands of DNA are now called chains and
the sites on the chain represent the nucleotides along the strand. The Schr�odinger equation
for a DNA molecule is written in form
(4.7)t(�1;n+1 + �1;n�1) + h�2;n = (E � "1;n)�1;n ;
t(�2;n+1 + �2;n�1) + h�1;n = (E � "2;n)�2;n ;
where t = 0:37 eV and h = 1:0 eV [99] are the intra-chain and inter-chain hopping
parameters, respectively. The on-site energies "i ;n are one of the four ionization energies of
the nucleotides (once again, "A = 8:24 eV, "T = 9:14 eV, "C = 8:87 eV and "G = 7:75 eV).
The bases A and T (C and G) are always found opposite of each other, A on one chain and
always T on the other one (same for C and G).
52
We employed the results of Chapter 3 (namely, Eq.(3.51) and Eq.(3.62)) to calculate the
inverse localization length numerically. We take E = 0 point equal to 8:5 eV, average of the
ionization energies of nucleotides. With that, the new on-site energies were "A = �0:26 eV,
"T = 0:64 eV, "C = 0:37 eV and "G = �0:75 eV. Similar to the calculations in Fishbone
model, I introduced a cut-o� value for k when we sum the �i j(k) terms. The summations
were run up to k < 400.
First, I compared the Fishbone and two-channel model results for BRCA-exon 11 region
and a segment of Human Chromosome 22 that includes both intron and exon regions. The
values of energy in the two calculations are not comparable, since the average energy E = 0
point is not same for the two di�erent models. In addition, the two-channel model results
(Fig.4.7 and Fig.4.8) include regions of energy where only one-channel is propagating and the
other channel is evanescent. I did not compare the two models as to where the delocalized
states occur in the energy spectrum, but whether the delocalized states exist in both models
for di�erent DNA regions. I found that exon region 11 of BRCA gene exhibits delocalized
electronic states in both of our calculations (Figs.4.5 and 4.7). On the other hand, the
studied segment of Chromosome 22 has no delocalized electronic states in the two-channel
model (Figs.4.6 and 4.8). Let me point to the di�erence in the on-site energy distributions
(binary disorder for Fishbone model, four valued disorder for two-channel model) as the main
reason behind the discrepancies between the two models. However it is not possible to
conclusively remark on why Chromosome 22 has only localized states by just investigating
the correlators. Following the discussion in Chapter 3, it can be said that oscillating and
slow decaying correlators of on-site energies are a necessary but not su�cient condition for
delocalized electron states in DNA.
4.4. Gene Expression and Quantum Electron Transport
I aimed to establish a link between the gene expression and the electron localization
length for a DNA molecule. The sequence of nucleotides determine the localization lengths
of electronic states of a given region. I asked myself whether a link between the functionality
of a segment of DNA and its electronic properties exists.
53
0
50
100
150
200
250
300
7 7.5 8 8.5 9 9.5 10
Loca
lizat
ion
Leng
th
Energy, eV
BRCA-Exon11
Figure 4.7. Localization length vs energy for BRCA gene exon region 11 is shown.
0
5
10
15
20
25
30
35
40
7 7.5 8 8.5 9 9.5 10
Loca
lizat
ion
Leng
th
Energy, eV
Chromosome 22 - Whole
Figure 4.8. Localization length vs energy for a segment of human Chromosome
22 DNA is shown. The segment includes both exon and intron regions.
In order to get an answer, I investigated the exon and intron regions of various genes
for their electron localization properties. The most evident di�erence in their functionality
54
0
50
100
150
200
7 7.5 8 8.5 9 9.5 10
Loca
lizat
ion
Leng
th
Energy, eV
SNAP29-Exon5
Figure 4.9. Localization length vs energy for human SNAP29 gene exon region
5 is shown.
is their role in protein coding. Two among the many DNA molecules we studied, human
SNAP29 and SYLB1 genes, are discussed below.
Fig.4.9 is the graph of localization length vs energy for SNAP29 exon region 5. The peaks
(l(E) > 200) at various energies represent the states with localization length exceeding the
length of the region signi�cantly. Those states are practically delocalized. Just like BRCA-
exon 11 and SYLB1-exon 7 regions, SNAP29-exon 5 region has a large number of states
that are delocalized. On the other hand, SNAP29-intron 2 and SYLB1- intron 2 regions and
the whole genes do not have any delocalized states. This is indeed the general trend for all
the DNA molecules I studied.
For the gene sequences, I have found a general rule of localization that apply :
� All exon regions have a number of delocalized states.
� All states in an intron region are localized.
� All states in the whole gene sequence are localized.
The �ndings di�erentiate exon and intron regions in existence of delocalized electronic
states. The non-existence of delocalized states for the whole (exon + intron) regions is a
55
0
5
10
15
20
25
30
35
40
7 7.5 8 8.5 9 9.5 10
Loca
lizat
ion
Leng
th
Energy, eV
SNAP29-Intron2
Figure 4.10. Localization length vs energy for human SNAP29 gene intron
region 2 is shown.
0
5
10
15
20
25
30
35
40
7 7.5 8 8.5 9 9.5 10
Loca
lizat
ion
Leng
th
Energy, eV
SNAP 29 - Whole
Figure 4.11. Localization length vs energy for human SNAP29 gene whole
sequence is shown.
consistency check since if the states in the intron regions are localized, they will localize
the states in the whole segment. This is a novel and striking result. It clearly shows that
56
0
50
100
150
200
7 7.5 8 8.5 9 9.5 10
Loca
lizat
ion
Leng
th
Energy, eV
SYLB1-Exon7
Figure 4.12. Localization length vs energy for human SYLB1 gene exon region
7 is shown.
whether a DNA region has or doesn't have delocalized electronic states may play a role in
its functionality, or vice versa. If we consider all the processes that take place during the
protein coding, it makes sense that electronic structure and an important aspect of it, the
localization length, has a determining role.
There are no distinct di�erences between the correlation functions of exon and intron
regions, so we do not have a direct means of understanding localization-delocalization from
the behavior of the correlation functions alone. Fig.4.15 and Fig.4.16 shows the correlation
function (only �11(k) is given in �gures) for SNAP29 Exon5 and Intron 2 regions, respectively.
57
0
10
20
30
40
50
60
70
80
7 7.5 8 8.5 9 9.5 10
Loca
lizat
ion
Leng
th
Energy, eV
SYLB1-Intron2
Figure 4.13. Localization length vs energy for human SYLB1 gene intron region
2 is shown.
0
5
10
15
20
25
30
35
40
45
7 7.5 8 8.5 9 9.5 10
Loca
lizat
ion
Leng
th
Energy, eV
SYLB1-Whole
Figure 4.14. Localization length vs energy for human SYLB1 gene whole se-
quence is shown.
58
-0.2
0
0.2
0.4
0.6
0.8
1
0 20 40 60 80 100
k
Correlation Function for SNAP29-Exon 5 Sequence
Figure 4.15. Correlation function �1(k) for SNAP29 gene exon region 5 is shown.
-0.2
0
0.2
0.4
0.6
0.8
1
0 20 40 60 80 100
k
Correlation Function of SNAP29-Intron 2 Sequence
Figure 4.16. Correlation function �1(k) for SNAP29 gene intron region 2 is shown.
59
CHAPTER 5
CONCLUSIONS
This thesis is a study of Anderson localization of electrons in one-dimensional two-channel
wires and in DNA molecules. It involves an analytical calculation part where we derive the
formula for the inverse localization length for electron states in a two-channel wire. It also
involves a computational part where the localization length is calculated for some DNA
molecules.
DNA is a one-dimensional wire. Other examples of a one-dimensial wire are carbon
nanotubes and wires etched on semiconducting �lms deposited on another semiconducting
device. Study of the localization of electronic states in multichannel wires is necessary to
understand the electron transport properties of 1-D wires.
This study of the localization length vs energy relationships in two-channel wires is the
�rst step in the generalization towards multichannel wires. Indeed, motivated with a similar
goal, Heinrichs [85] has found the formula for the inverse localization length vs energy for
two-channel and three-channel disordered wires with uncorrelated weak disorder. I generalized
the formula for two-channel wires with correlated weak disorder [103].
One technologically interesting application of the results of this thesis has to do with
carbon nanotubes. Nanotubes are graphene layers wrapped into a cylinder. The diameter
and chirality of nanotubes determine whether they are metallic or semiconducting [105]. In
practice most nanotubes are quasi-1D wires with a few number of conducting channels.
The disorder in nanotubes can be of almost any form, depending on the production, deposi-
tion/doping or the environment. Most molecules and atoms, including metal atoms/structures
attach to nanotubes without altering its electronic state structure signi�cantly [104]. In those
cases, nanotubes are weakly disordered systems. I believe this study in localization will be
important for various technological applications of nanotubes.
60
Yet another important �eld of application is the DNA molecules. My work on electron
localization in single-channel Fishbone model for DNA molecules show that the long-range
sequential correlations in DNA may in fact lead to electron delocalization [102].
My study in localization of electron states in DNA as a two-channel wire has very striking
and far reaching results. First of all, our study shows that DNA molecules can have localized
or delocalized electron states depending on their genetic sequence information. Second and
more importantly, the calculations give a link between the functionality of DNA regions and
their electron localization properties. The general rule of localization itemized in Chapter
4 states that all electron states in intron regions are localized and all exon regions have a
number of delocalized states.
At the moment, I do not make any claims as to how or why such a di�erence in localization
of electronic states exists between intron and exon regions. I simply point out at the statistical
information available to the scienti�c community. The vast information on sequences of genes
of various organisms can be very quickly analyzed by just looking at the correlations in the
sequence disorder. It still remains a laborious task to analyze the sequences in relation to
each other, going step by step from chromosome to chromosome, organism to organism, but
we propose a possible starting point.
Long sequences of DNA may be mapped according to their electron localization properties
using a sliding window technique. The segment of DNA inside the window will be characterized
by its electron localization length that can be easily calculated from the correlation function
for the sequence of the segment. This will create a sort of chart that can be used as reference
or as a �ngerprint. The chart will point out to regions of certain electronic properties that
may be desirable for various applications.
Another important point of investigation for near future is to see whether small changes
in the sequence leads to any change in the localization length. Cancerous DNA segments
can be cross-checked with healthy DNA to see if exon regions of cancerous genes show any
discrepancy from the general rule of localization.
61
Finally, I believe that the studies parallel to this work for DNA segments may give some
information about how various regions migrate from generation to generation and what role
some parts of DNA may have played in the process of evolution.
62
REFERENCES
[1] J. D. Watson and F. H. C. Crick, Nature 171 (1953),737.
[2] C.-K. Peng, S. V. Buldyrev, A. L. Goldberger, S. Havlin, F. Sciortino, M. Simons, and H. E. Stanley,
Nature 356 (1992), 168.
[3] R. F. Voss, Phys. Rev. Lett. 68 (1992), 3805.
[4] S. V. Buldyrev, A. L. Goldberger, S. Havlin, C.-K. Peng, M. Simons, and H. E. Stanley, Phys. Rev. E 47
(1993), 4514.
[5] P. Allegrini, M. Barbi, P. Grigolini, and B. J. West, Phys Rev E 52 (1995),5281.
[6] N. Scafetta, V. Latora, and P. Grigolini, Phys Rev E 66 (2002),031906.
[7] P. J. Dandliker, R. E. Holmlin, and J. K. Barton, Science 275 (1997), 1465.
[8] J. Hihath, B. Xu, P. Zhang, and N. Tao, Proc. National Acad. Sc. 102 (2005), 16979.
[9] R. G. Endres, D. L. Cox, and R. P. Singh, Rev. Mod. Phys. 76 (2004), 195214.
[10] M. Mertig et al., Eur. Phys. J. D 9 (1999), 45.
[11] K. Keren et al., Science 302 (2003), 1380.
[12] E. Braun, Y. Eichen, U. Sivan, and G. Ben-Yoseph, Nature 391 (1998),775778 ; Y. A. Berlin, A. L. Burin,
and M. A. Ratner, J Phys Chem A 104 (2000), 443.
[13] D. Porath, A. Bezryadin, S. Vries, and C. Dekker, Nature 403 (2000), 635638.
[14] H. V. Fink and C. Schonenberger, Nature 398 (1999), 407410.
[15] M. A. Ratner, Nature 397 (1999), 480.
[16] P. W. Anderson, Phys. Rev. 109 (1958), 14921505.
[17] D. J. Thouless, Phys. Rev. Lett. 61 (1998), 2141.
[18] F. M. Izrailev and A. A. Krokhin, Phys. Rev. Lett. 82 (1999), 4062.
[19] F. M. Izrailev, A. A. Krokhin, and S. Ulloa S, Phys. Rev. B 63 (2001), 041102.
[20] D. Matteo, Ultramicroscopy Volume 107 (2007), 184-190.
[21] R. S. Fellers, D. Barsky, F. Gygi and M. Colvin, Chem. Phys. Lett. 312 (1999), 548-555(8).
[22] E. Caut, D. Dehareng, and J. Livin, J. Phys. Chem. A 110 (2006), 92009211.
[23] D. Roca-Sanjun, M. Rubio, M. Merchn, and L. Serrano-Andrs, J. Chem. Phys. 125(2006), 084302-084302-
7.
63
[24] G. M. Cooper, The Cell, a Molecular Approach (ASM Press and Sinauer Associates, Inc.), Washington,
D. C., 2000.
[25] Y. Imry, Introduction to Mesoscopic Physics (Oxford University Press), New York, 1997.
[26] A. Wiessmann et. al., Phys. Rev. Lett. 38 (1977), 782.
[27] J. E. Mooil, Phys. Stat. Sol. A17 (1973), 521.
[28] L. P. Gorkov, A. Larkin, and I. Khmelnitskii, JTEP Lett. 30 (1979), 288.
[29] P. A. Mello and N. Kumar, Quantum Transport in Mesoscopic Systems (Oxford University Press), Newyork.
(2004)
[30] N. F. Mott, J. Non-Crysr. Solids 1 (1968), 1.
[31] D. J. Thouless, Phys. Rep. 13 (1974), 93.
[32] B. B. Schmidt, M. H. Hettler, and G. Schn, Phys. Rev.B 75 (2007), 115125.
[33] H. Mehrez and M. P. Anantram, Phys. Rev. B 71 (2005), 115405.
[34] A.V. Malyshev, Phys. Rev. Lett. 98 (2007), 096801.
[35] E. Maci, Phys. Rev. B 74 (2006), 245105.
[36] H. A. Wagenknecht (Editor), Charge Transfer in DNA: From Mechanism to Application, ISBN: 978-3-
527-31085-2, August 2005.
[37] Y. J. Yan and H. Zhang, J. Theor. Comput. Chem. 1 (2002), 225.
[38] B. Kramer and A. MacKinnon, Rep. Prog. Phys. 56 (1993), 14691564.
[39] N. F. Mott and E. A. Davis, Electronic Processes in Non-Crymlline Malerids 2nd edn (Oxford: Clarendon),
1979.
[40] E. Abrahnms, P. W. Anderson, D. C. Licciardello and T. V. Ramakrishnan, Phhs. Rev. Lett. 42 (1979),
613.
[41] D. J. Thouless, Phys. Rev. Lett. 39 (1977), 1167.
[42] R. Johnston and H. Kunz, Jour. Phys. C 16 (1983), 3895.
[43] P. E. Wolf and G. Maret, Phys. Rev. Lett. 55 (2005), 2696.
[44] D. Yu Shan'in and V. Y. Sharvin, JETP Lett. 34 (1981), 272.
[45] P. Sheng, Scattering and Localization of Classical Waves in Random Media (Singapore: World Scientic)
1990.
[46] Y. Kuga and A. lshimaru, J. 0pt. Soc. Am. A 1 (1984), 1831.
[47] N. F. Mott and D. W. Twose, Adv. Phys. 10 (1961), 107-163.
[48] R. E. Borland, Proc. Roy. Soc. A 274 (1963), 529-545.
[49] B. Y. Tong, Phys. Rev. A 1 (1970), 52-58.
[50] K. Ishii, Prog. Theor. Phys. Suppl. 53 (1974), 77-138.
64
[51] M. Griniasty and S. Fishman, Phys. Rev. Lett. 60 (1988), 1334-1337.
[52] S. D. Sarma, S. He, and X. C. Xie, Phys. Rev. Lett. 61 (1988), 2144-2147.
[53] D. H. Dunlap, H-L. Wu, and P. W. Phillips, Phys. Rev. Lett. 65 (1990), 88-91.
[54] J. C. Flores, J. Phys.: Condensed Matter 1 (1989), 8471-8479.
[55] P. Phillips and H-L. Wu, Science 252 (1991), 1805-1812.
[56] A. Bovier, J. Phys. A.:Math. Gen. 25 (1992), 1021-1029.
[57] J. C. Flores and M. Hilke, J. Phys. A : Math. Gen. 26 (1993), L1255-L1259.
[58] F. M. Izrailev, T. Kottos and G. P. Tsironis, J. Phys.: Condensed Matter 8 (1996), 2823-2834.
[59] A. S�anchez, F. Dom��nguez-Adame, G. Berman F. M. Izrailev, Phys. Rev. B 51 (1995), 6769-6772.
[60] T. Sedrakyan, Phys. Rev. B 69 (2004), 085109.
[61] E. Diez, A. S�anchez and F. Dom��nguez-Adame,IEEE Journal of Quantum Electronics 31 (1995), 1919-
1926.
[62] E. Diez, A. S�anchez and F. Dom��nguez-Adame, Phys. Rev. B 50 (1994), 17736-17739.
[63] M. Hilke, J. Phys.: Math. Gen. 30 (1997), L367-L371.
[64] E. Lazo and M. E. Onell, Phys. Lett. A 283 (2001), 376-381.
[65] E. Lazo and M. E. Onell, Physica B 299 (2001), 173-179.
[66] X. Q. Huang, R. W. Peng, F. Qiu, S. S. Jiang and A. Hu, Eur. Phys. J. B 23 (2001), 275-281.
[67] F. M. Izrailev, T. Kottos and G. P. Tsironis, Phys. Rev. B 52 (1995), 3274-3279.
[68] T. Kottos and G. P. Tsironis and F. M. Izrailev, J. Phys.: Condensed Matter 9 (1997), 1777-1791.
[69] L. Tessieri and F. M. Izrailev, Physica E 9 (2001), 405-417.
[70] W. Zhang and S. E. Ulloa, Phys. Rev. B 69 (2004), 153203.
[71] B. Derrida and E. Gardner, J. Physique 45 (1984), 1283.
[72] F.M. Izrailev, S. Ru�o, and L. Tessieri, J. Phys.: Math. Gen. 31 (1998), 5263.
[73] L.I. Deych, A.A. Lisyansky, and B.L. Alstshuler, Phys. Rev. Lett. 84 (2000), 2678.
[74] L.I. Deych, M.V. Erementchouk, and A.A. Lisyansky, Phys. Rev. Lett. 90 (2003), 126601.
[75] J.M. Luck, Phys. Rev. B 39 (1989), 5834.
[76] H. Schomerus and M. Titov, Phys. Rev. B 67 (2003), 100201(R).
[77] T. Sedrakyan and A. Ossipov, Phys. Rev. B 70 (2004), 214206.
[78] R. A. Caetano and P. A. Schulz, Phys. Rev. Lett. 95 (2005), 126601.
[79] E. D��az, A. Sedrakyan, D. Sedrakyan, and F. Dom��nguez-Adame, Phys. Rev. B 75 (2007), 014201.
[80] F. A. B. F. de Moura and M. L. Lyra, Phys. Rev. Lett. 81 (1998), 3735-3738.
[81] F. A. B. F. de Moura and M. L. Lyra, Physica A 266 (1999), 465-470.
[82] L. Tessieri, J. Phys. A: Math. Gen. 35 (2002), 9585-9600.
65
[83] S. Russ, J. W. Kantelhardt, A. Bunde and S. Havlin, Phys. Rev. B 64 (2001), 134209.
[84] H. Shima, T. Nomura and T. Nakayama, Phys. Rev. B 70 (2004), 075116.
[85] J. Heinrichs, Phys. Rev. B 66 (2002), 155434.
[86] J. Heinrichs, Phys. Rev. B 68 (2003), 155403.
[87] U. Kuhl, F. M. Izrailev, A. A. Krokhin, H.J. St�ockmann, Appl. Phys. Lett. 77 (2000), 633.
[88] Y. Imry and R. Landauer, Rev. Mod. Phys. 71 (1999), S306.
[89] S. Datta, Electronic Transport in Mesoscopic Systems (Cambridge University Press), 1995.
[90] B. J. van Wees, H. Van Houten, C. W. J. Beenakker, J. G. Williamson, L. P. Kouwenhoven, D. Van der
Marel, and C. T. Foxon, Phys. Rev. Lett. 60 (1988), 848.
[91] H. Van Houten and C. W. J. Beenakker, Phys. Today 49(7) (1996), 22.
[92] A. Crisanti,G. Paladin and A. Vulpiani, Products of random matrices in statistical physics (Springer-Verlag),
1993.
[93] V. I. Oseledec, Trans. Mosc. Math. Soc. 19 (1968), 197.
[94] V. N.Tutubalin, Theor. Probab. Appl. 13 (1968), 65.
[95] M. Janssen, Phys. Rep. 295 (1998), 1.
[96] P. W. Anderson and P. A. Lee, Suppl. Prog. Theor. Phys. 69 (1980), 212.
[97] E. Mac��a and S. Roche, Nanotechnology 17 (2006) 30023007.
[98] G. Cuniberti, L. Craco, D. Porath and C. Dekker, Phys. Rev. B 65 (2002), 241314.
[99] D. Klotsa, R. A. R�omer and M. S. Turner, Biophys J 89 (2005), 218798.
[100] W. Zhang, S. E. Ulloa, Phys. Rev. B 69 (2004), 153203.
[101] GenBank Database. Available from www.ncbi.nlm.nih.gov/Genbank.
[102] V. M. K. Bagci and A. A. Krokhin, Chaos, Solitons and Fractals 34 (2007) 104111.
[103] V. M. K. Bagci and A. A. Krokhin, Phys. Rev. B 76 (2007), 134202.
[104] V. M. K. Bagci et. al., Phys. Rev. B 67, 201401 (2003); V. M. K. Bagci et. al., Phys. Rev. B 66, 045409
(2003)
[105] http://www.lancs.ac.uk/users/esqn/nanoelectronics/presentations/gueron.pdf
[106] Y. Chen, X. Shi-Jie and S. N. Evangelou, J. Phys.: Condens. Matter 10 (1998), 1065-1077.
[107] Y. Imry, Europhys. Lett. 1 (1986), 249-256.
[108] S. Kurachi, Y. Hitomi, M. Furukawa and K. Kurachi, J. Biological Chemistry 270 (1995), 5276-5281.
66